Handbook

Haynes 4th proof 13 December 2017 13/12/2017 15:37 Page i
Metadata for
Information
Management and
Retrieval
Haynes 4th proof 13 December 2017 13/12/2017 15:37 Page ii
Every purchase of a Facet book helps to fund CILIP’s advocacy,

awareness and accreditation programmes for
information professionals.
Haynes 4th proof 13 December 2017 13/12/2017 15:37 Page iii
Metadata for
Information
Management and
Retrieval
Understanding metadata and its use
Second edition
David Haynes
Haynes 4th proof 13 December 2017 13/12/2017 15:37 Page iv
© David Haynes 2004, 2018
Published by Facet Publishing

7 Ridgmount Street, London WC1E 7AE
www.facetpublishing.co.uk
Facet Publishing is wholly owned by CILIP: the Library and Information

Association.
The author has asserted his right under the Copyright, Designs and Patents Act
1988 to be identified as author of this work.
Except as otherwise permitted under the Copyright, Designs and Patents Act
1988 this publication may only be reproduced, stored or transmitted in any form
or by any means, with the prior permission of the publisher, or, in the case of
reprographic reproduction, in accordance with the terms of a licence issued by
The Copyright Licensing Agency. Enquiries concerning reproduction outside
those terms should be sent to Facet Publishing, 7 Ridgmount Street, London
WC1E 7AE.
Every effort has been made to contact the holders of copyright material
reproduced in this text, and thanks are due to them for permission to reproduce
the material indicated. If there are any queries please contact the publisher.
British Library Cataloguing in Publication Data
A catalogue record for this book is available from the British Library.
ISBN 978-1-85604-824-8 (paperback)

ISBN 978-1-78330-115-7 (hardback)
ISBN 978-1-78330-216-1 (e-book)
First published 2004

This second edition, 2018
Text printed on FSC accredited material.
Typeset from author’s files in 10/13 pt Palatino Lintoype and Open Sans by
Flagholme Publishing Services.
Printed and made in Great Britain by CPI Group (UK) Ltd, Croydon, CR0 4YY.
Haynes 4th proof 13 December 2017 13/12/2017 15:37 Page v
Contents
List of figures and tables ix

Preface xi
Acknowledgements xiii
PART I METADATA CONCEPTS 1

1 Introduction 3
Overview 3
Why metadata? 3
Fundamental principles of metadata 4
Purposes of metadata 11
Why is metadata important? 17
Organisation of the book 17
2 Defining, describing and expressing metadata 19

Overview 19
Defining metadata 19
XML schemas 24
Databases of metadata 26
Examples of metadata in use 27
Conclusion 33
3 Data modelling 35
Overview 35
Metadata models 35
Unified Modelling Language (UML) 36
Resource Description Framework (RDF) 36
Dublin Core 39
The Library Reference Model (LRM) and the development of RDA 40
ABC ontology and the semantic web 42
Indecs – Modelling book trade data 44
Haynes 4th proof 13 December 2017 13/12/2017 15:37 Page vi
VI METADATA FOR INFORMATION MANAGEMENT AND RETRIEVAL
OAIS – Online exchange of data 46

Conclusion 48
4 Metadata standards 49
Overview 49
The nature of metadata standards 49
About standards 51
Dublin Core – a general-purpose standard 51
Metadata standards in library and information work 54
Social media 62
Non-textual materials 64
Complex objects 70
Conclusion 74
PART II PURPOSES OF METADATA 75

5 Resource identification and description (Purpose 1) 77
Overview 77
How do you identify a resource? 77
Identifiers 78
RFIDs and identification 85
Describing resources 86
Descriptive metadata 88
Conclusion 93
6 Retrieving information (Purpose 2) 95

Overview 95
The role of metadata in information retrieval 95
Information Theory 97
Types of information retrieval 98
Evaluating retrieval performance 102
Retrieval on the internet 104
Subject indexing and retrieval 106
Metadata and computational models of retrieval 107
Conclusion 111
7 Managing information resources (Purpose 3) 113

Overview 113
Information lifecycles 113
Create or ingest 117
Preserve and store 118
Distribute and use 122
Review and dispose 123
Transform 124
Conclusion 124
8 Managing intellectual property rights (Purpose 4) 127

Overview 127
Rights management 127
Haynes 4th proof 13 December 2017 13/12/2017 15:37 Page vii
CONTENTS VII
Provenance 134
Conclusion 137
9 Supporting e-commerce and e-government (Purpose 5) 139

Overview 139
Electronic transactions 139
E-commerce 140
Online behavioural advertising 141
Indecs and ONIX 143
Publishing and the book trade 144
E-government 148
Conclusion 149
10 Information governance (Purpose 6) 151

Overview 151
Governance and risk 151
Information governance 153
Compliance (freedom of information and data protection) 154
E-discovery (legal admissibility) 156
Information risk, information security and disaster recovery 156
Sectoral compliance 158
Conclusion 159
PART III MANAGING METADATA 161

11 Managing metadata 163
Overview 163
Metadata is an information resource 163
Workflow and metadata lifecycle 164
Project approach 165
Application profiles 170
Interoperability of metadata 171
Quality considerations 179
Metadata security 181
Conclusion 182
12 Taxonomies and encoding schemes 185

Overview 185
Role of taxonomies in metadata 185
Encoding and maintenance of controlled vocabularies 186
Thesauri and taxonomies 188
Content rules – authority files 191
Ontologies 194
Social tagging and folksonomies 199
Conclusion 201
13 Very large data collections 203

Overview 203
The move towards big data 203
Haynes 4th proof 13 December 2017 13/12/2017 15:37 Page viii
VIII METADATA FOR INFORMATION MANAGEMENT AND RETRIEVAL
What is big data? 205

The role of linked data in open data repositories 206
Data in an organisational context 209
Social media, web transactions and online behavioural 211
advertising
Research data collections 212
Conclusion 219
14 Politics and ethics of metadata 221

Overview 221
Ethics 221
Power 226
Money 229
Re-examining the purposes of metadata 230
Managing metadata itself 236
Conclusion 237
References 239
Index 257
Haynes 4th proof 13 December 2017 13/12/2017 15:37 Page ix
List of figures and tables
Figures
1.1 Metadata from the Library of Congress home page 12
2.1 Example of marked-up text 20
2.2 Rendered text 21
2.3 Word document metadata 28
2.4 Westminster Libraries – catalogue search 30
2.5 Westminster Libraries catalogue record 30
2.6 WorldCat search 31
2.7 WorldCat detailed record 32
2.8 OpenDOAR search of repositories 32
2.9 Detailed OpenDOAR record 33
3.1 An RDF triple 37
3.2 More complex RDF triple 37
3.3 A triple expressed as linked data 38
3.4 DCMI resource model 39
3.5 Relationships between Work, Expression, Manifestation and Item 41
3.6 LRM agent relationships 42
3.7 Publication details using the ABC Ontology 44
3.8 Indecs model 45
3.9 OAIS simple model 46
3.10 OAIS Information Package 46
3.11 Relationship between Information Packages in OAIS 47
4.1 BIBFRAME 2.0 model 57
4.2 Overlap between image metadata formats 66
4.3 IIIF object 67
4.4 Relationships between IIIF objects 67
4.5 Metadata into an institutional repository 72
4.6 How OAI-PMH works 72
5.1 Example of relationship between ISTC and ISBN 85
5.2 Structure of an Archival Resource Key 85
Haynes 4th proof 13 December 2017 13/12/2017 15:37 Page x
X METADATA FOR INFORMATION MANAGEMENT AND RETRIEVAL
6.1 Resolution power of keywords 96

6.2 Boolean operators 100
6.3 British Library search interface 108
6.4 Metadata fields in iStockphoto 111
7.1 DCC simplified information lifecycle 116
7.2 Generic model of information lifecycle 116
7.3 PREMIS data model 121
7.4 Loan record from Westminster Public Libraries 123
8.1 ODRL Foundation Model 131
8.2 Legal view of entities in ONIX 132
8.3 Creative Commons Licence 133
8.4 PROV metadata model for provenance 135
9.1 Cookie activity during a browsing session 142
9.2 ONIX e-commerce transactions 146
11.1 Stages in the lifecycle of a metadata project 166
11.2 Singapore Framework 170
11.3 Possible crosswalks between four schemas 177
11.4 Possible crosswalks between ten schemas 177
11.5 Data Catalog Vocabulary Data Model 178
11.6 A-Core Model 180
12.1 Extract from an authority file from the Library of Congress 192
12.2 Conceptual model for authority data 192
12.3 Use of terms from a thesaurus 193
12.4 Google Knowledge Graph results 197
12.5 Structured data in Google about the British Museum 198
13.1 Screenshot of search results from the European Data Portal 208
13.2 Agents involved in delivering online ads to users 212
13.3 A ‘pyramid’ of requirements for reusable data 214
13.4 Silo-based searching 218
13.5 Federated search service 218
13.6 Index-based discovery system 219
Tables
1.1 Day’s model of metadata purposes 13
1.2 Different types of metadata and their functions 14
4.1 KBART fields 60
4.2 IIIF resource structure 68
11.1 Dublin Core to MODS Crosswalk 176
13.1 Comparison of metadata fields required for data sets in Project Open Data 209
13.2 Core metadata elements to be provided by content providers 213
14.1 Metadata standards development 231
Haynes 4th proof 13 December 2017 13/12/2017 15:37 Page xi
Preface
T
HIS IS NOT A ‘HOW TO DO IT’ BOOK. There are several excellent guides
about the practical steps for creating and managing metadata. This
book is intended as a tutorial on metadata and arose from my own
need to find out more about how metadata worked and its uses. The original
book came out at a time when there were very few guides of this type
available. Metadata Fundamentals for All Librarians provided a good starting
point which introduced the basic concepts and identified some of the main
standards that were then available (Caplan, 2003). It was an early publication
from a period of tremendous development and in an area that was changing
day to day. Introduction to Metadata, published by the Getty Institute,
represented another milestone and provided more comprehensive
background to metadata (Baca, 1998). It is now in its third edition (Baca, 2016).
In my work as an information management consultant many colleagues
and clients kept asking the questions: ‘What is metadata?’, ‘How does it
work?’, and ‘What’s it for?’. The last of these questions particularly resonated
with the analysis and review of information services. This led to the
development of a view of metadata defined by its purposes or uses. Since the
first edition of Metadata for Information Management and Retrieval there have
been many excellent additions to the literature, notably Zeng and Qin’s book,
simply entitled Metadata, which is now in its second edition (Zeng and Qin,
2008; 2015; Haynes, 2004). I also enjoyed Philip Hider’s book, Information
Resource Description, which is substantially about metadata from a subject
retrieval perspective (Hider, 2012). There are many other excellent tomes,
some of which are mentioned in the main body of this book. I hope that this
second edition adds a unique perspective to this burgeoning field.
Haynes 4th proof 13 December 2017 13/12/2017 15:37 Page xii
XII METADATA FOR INFORMATION MANAGEMENT AND RETRIEVAL
This book covers the basic concepts of metadata and some of the models
that are used for describing and handling it. The main purpose of this book
is to reveal how metadata operates, from the perspective of the user and the
manager. It is primarily concerned with data about document-based
information content – in the broadest sense. Many of the examples will be for
bibliographic materials such as books, e-journals and journal articles.
However, this book also covers metadata about the documentation associated
with museum objects (thus making them information objects), as well as
digital resources such as research data collections, web resources, digitised
images, digital photographs, electronic records, music, sound recordings and
moving images. It is not a book about databases or data modelling, which is
covered elsewhere (Hay, 2006).
Metadata for Information Management and Retrieval is international in
coverage and sets out to introduce the concepts behind metadata. It focuses
on the ways metadata is used to manage and retrieve information. It
discusses the role of metadata in information governance as well as exploring
its use in the context of social media, linked open data and big data. The book
is intended for museums, libraries, archives and records management
professionals, including academic libraries, publishers, and managers of
institutional repositories and research data sets. It will be directly relevant to
students in the iSchools as well as those who are preparing to work in the
library and information professions. It will be of particular interest to the
knowledge organisation and information architecture communities. Managers
of corporate information resources and informed users who need to know
about metadata will also find much that is relevant to them. Finally, this book
is for researchers who deal with large data sets, either as their creators or as
users who need to understand the ways in which that data is described, its
properties and ways of handling and interrogating that data.
David Haynes, August 2017

Haynes 4th proof 13 December 2017 13/12/2017 15:37 Page xiii
Acknowledgements
P
REPARATION OF THIS BOOK would not have been possible without the
support and assistance of many individuals, too numerous to list. I
hope that they will recognise their contributions in this book and will
accept this acknowledgement as thanks. Any shortcomings are entirely my
own.
I would like to thank colleagues at City, University of London. David
Bawden and Lyn Robinson at the Centre for Information Science provided
guidance and encouragement throughout. Andy MacFarlane was an excellent
critic for the early drafts of the chapter on information retrieval. The library
service at City, University of London has been an invaluable resource which,
with the back-up of the British Library, has been essential for the identification
and procurement of relevant literature.
Neil Wilson, Rachael Kotarski, Bill Stockting and Paul Clements at the
British Library, Christopher Hilton at the Wellcome Library and Graham Bell
of EDItEUR all freely gave their time in interviews and follow-up questions.
I would like to acknowledge the contribution made by former colleagues
at CILIP, where I was working when I wrote the first edition. I am also
grateful for the feedback from reviewers, colleagues and students who have
used the book as a text. I am especially grateful for the moral support of the
University of Dundee, where I teach a module on ‘Metadata Standards and
Information Taxonomies’ on their postgraduate course in the Centre for
Archives and Information Studies (CAIS). Teaching that particular course has
helped to shape my thinking and has given me an incentive to read and think
more about metadata.
Many colleagues in the wider library and information profession helped to
clarify specific points about the use of metadata. I would especially like to
Haynes 4th proof 13 December 2017 13/12/2017 15:37 Page xiv
XIV METADATA FOR INFORMATION MANAGEMENT AND RETRIEVAL
thank Gordon Dunsire for going through the manuscript and pointing out
significant issues that I hope have now been addressed.
Finally I would like to thank family, friends and colleagues who have
provided constant encouragement throughout this enterprise.
Haynes 4th proof 13 December 2017 13/12/2017 15:37 Page 1
PART I
Metadata concepts
Part I introduces the concepts that underpin metadata, starting with an
historical perspective. Some examples of metadata that people come across
in their daily life are demonstrated in Chapter 1, along with some alternative
views of metadata and how it might be categorised. This chapter defines the
scope of this book as considering metadata in the context of document
description. Chapter 2 looks at mark-up languages and the development of
schemas as a way of representing metadata standards. It also highlights the
connection between metadata and cataloguing. Chapter 3 looks at different
ways of modelling data with specific reference to the Resource Description
Framework (RDF). It describes the Library Reference Model (LRM) and its
impact on current cataloguing systems. Chapter 4 discusses cataloguing and
metadata standards and ways of representing metadata. It introduces RDA,
MARC, BIBFRAME as well as standards used in records management, digital
repositories and non-textual materials such as images, video and sound.
CHAPTER 1
Introduction
Overview
This chapter sets out to introduce the concepts behind metadata and illustrate them with
historical examples of metadata use. Some of these uses predate the term ‘metadata’. The
development of metadata is placed in the context of the history of cataloguing, as well as
parallel developments in other disciplines. Indeed, one of the ideas behind this book is that
metadata and cataloguing are strongly related and that there is considerable overlap
between the two. Pomerantz (2015) and Gartner (2016) have made a similar connection,
although Zeng and Qin (2015) emphasise the distinction between cataloguing and
metadata. This leads to discussion of the definitions of ‘metadata’ and a suggested form
of words that is appropriate for this book. Examples of metadata use in e-publishing,
libraries, archives and research data collections are used to illustrate the concept. The
chapter then considers why metadata is important in the wider digital environment and
some of the political issues that arise. This approach provides a way of assessing the
models of metadata in terms of its use and its management. The chapter finally introduces
the idea that metadata can be viewed in terms of the purposes to which it is put.
Why metadata?
If anyone wondered about the importance of metadata, the Snowden
revelations about US government data-gathering activities should leave no
one in any doubt. Stuart Baker, the NSA (National Security Agency) General
Counsel, said ‘Metadata tells you everything about somebody’s life. If you
have enough metadata you don’t really need content’ (Schneier, 2015, 23). The
routine gathering of metadata about telephone calls originating outside the
4 PART I METADATA CONCEPTS
USA or calls to foreign countries from the USA caused a great deal of concern,
not only among American citizens but also among the US’s strongest allies and
trading partners. The UK’s Investigatory Powers Act (UK Parliament, 2016)
requires communications providers to keep metadata records of commun-
ications via public networks (including the postal network) to facilitate security
surveillance and criminal investigations. As Jacob Appelbaum said when the
Wikileaks controversy first blew up, ‘Metadata in aggregate is content’
(Democracy Now, 2013). His point was that when metadata from different
sources is aggregated it can be used to reconstruct the information content of
communications that have taken place.
Although metadata has only recently become a topic for public discussion,
it pervades our lives in many ways. Anyone who uses a library catalogue is
dealing with metadata. Since the first edition of this book the idea of metadata
librarians or even metadata managers has gained traction. Job advertisements
often focus on making digital resources available to users. Roles that would
have previously been described in terms of cataloguing and indexing are
being expressed in the language of metadata. Re-use of data depends on
metadata standards that allow different data sources to be linked to provide
innovative new services. Many apps on mobile devices depend on combining
location with live data feeds for transportation, air quality or property prices,
for example. They depend on metadata.
Fundamental principles of metadata

Some historical background
Although the term ‘metadata’ is a recent one, many of the concepts and
techniques of metadata creation, management and use originated with the
development of library catalogues. If we regard books and scrolls as
information objects, a book catalogue could be seen to be a collection of
metadata. It contains data about information objects. An understanding of
what people tried to do before the term ‘metadata’ was coined helps to
explain the concept of metadata. The historical background also gives a
perspective on why metadata has become so important in recent years.
The idea of cataloguing information has been around at least since the
Alexandrian Library in ancient Egypt. Callimachus of Cyrene (305–235 BC),
the poet and author, was a librarian at Alexandria. He is widely credited with
creating the first catalogue, the Pinakes, of the Alexandrian Library’s 500,000
scrolls. The catalogue was itself a work of 120 scrolls with titles grouped by
subject and genre. This could be seen as the first recorded compilation of
metadata. Gartner (2016) provides an elegant description of the history of
metadata from antiquity to the present.
INTRODUCTION 5
In Western Europe library cataloguing developed in the ecclesiastical and,

later, academic libraries. In the eighth century AD the books donated by
Gregory the Great to the Church of St Clement in Rome were catalogued in
the form of a prayer. During the same era, Alcuin of York (735–804) developed
a metrical catalogue for the cathedral library at York. Cataloguing developed,
so that by the 14th century the location of books started to appear in catalogue
records and by the 16th century the first alphabetical arrangements began to
appear. Up until that time catalogues were used as inventories of stock rather
than for finding books or for managing collections.
Modern library catalogues date back to the French code of 1791, the first
national cataloguing code with author entry, which used catalogue cards and
rules of accessioning and guiding. Cataloguing rules (an important aspect of
metadata) were developed by Sir Anthony Panizzi for the British Museum
Library and these were published in 1841. In the USA Charles A. Cutter
prepared Rules of a Dictionary Catalog, which was published in 1876. The
American Library Association and the Library Association in the UK both
developed cataloguing rules around the start of the 20th century. This led to
an agreement in 1904 to co-operate to produce an international cataloguing
code, which was published as separate American and British editions in 1908.
Later, the International Conference on Cataloguing Principles in Paris in
1961 established a set of principles on the choice and form of headings in
author/title catalogues. These were incorporated into the first edition of the
Anglo-American Cataloguing Rules (AACR) in 1967, published in two
versions by the Library Association and the American Library Association
(Joint Steering Committee for Revision of AACR & CILIP, 2002).The
International Standard Bibliographic Descriptions (ISBDs) were developed
by IFLA, the International Federation of Library Associations, and were
incorporated into the second edition of the Anglo-American Cataloguing
Rules (AACR2), published in 1978. ISBD specifies the sources of information
used to describe a publication, the order in which the data elements appear
and the punctuation used to separate the elements. Material-specific ISBDs
were merged into a consolidated edition (IFLA, 2011). AACR2 specifies how
the values of the data elements are determined. This was an important
development because it made catalogues more interchangeable and allowed
for conversion into machine-readable form (Bowman, 2003).
In the mid-1960s computers started being used for the purpose of
cataloguing and a new standard for the data format of catalogue records,
MARC (Machine Readable Cataloguing) was established. MARC covers all
kinds of library materials and is usable in automated library management
systems. Although MARC was initially used to process and generate
catalogue cards more quickly, libraries soon started to use this as a means of
exchanging cataloguing data, which helped to reduce the cost of cataloguing

original materials. The availability of MARC records stimulated the
development of searchable electronic catalogues. The user benefited from
wider access to searchable catalogues, and later on to union catalogues, which
allowed them to search several library catalogues at once. Different versions
of MARC emerged, largely based on national variations e.g. USMARC,
UKMARC and Norway’s NORMARC. Although the different MARC
versions were designed to reflect the particular needs and interests of different
countries or communities of interest, this inhibited international exchange of
records. It was only with the widespread adoption of MARC 21 by the
national bibliographic authorities that a degree of harmonisation of national
bibliographies was achieved.
The growth of electronic catalogues and the development of textual
databases able to handle summaries of published articles demanded new
skills, which in turn contributed to the development of information science
as a discipline. Information scientists developed many of the early electronic
catalogues and bibliographic databases (Feather and Sturges, 1997). They
adapted library cataloguing rules for an electronic environment and did much
of the pioneering work on information retrieval theory, including the
measures of precision and recall which are discussed in Chapter 6.
Although metadata was first used in library catalogues it is now widely
used in records management, the publishing industry, the recording industry,
government, the geospatial community and among statisticians. Its success
as an approach may be because it provides the tools to describe electronic
information resources, allowing for more consistent retrieval, better
management of data sources and exchange of data records between
applications and organisations.
Vellucci (1998) suggested that the term ‘metadata’ dates back to the 1960s
but became established in the context of Database Management Systems
(DBMS) in the 1970s. The first reference to ‘meta-data’ can be traced back to
a PhD dissertation, ‘An infological approach to data bases’, which made the
distinction between (Sundgren 1973):
• objects (real-world phenomena)

• information about the object
• data representing information about the object (i.e. meta-data).
The term began to be widely used in the database research community by the
mid-1970s.
A parallel development occurred in the geographical information systems
(GIS) community and in particular the digital spatial information discipline.
INTRODUCTION 7
In the late 1980s and early 1990s there was considerable activity within the
GIS community to develop metadata standards to encourage interoperability
between systems. Because government (especially local government) activity
often requires data to describe location, there are significant benefits to be
gained from a standard to describe location or spatial position across
databases and agencies. The metadata associated with location data has
allowed organisations to maintain their often considerable internal
investments in geospatial data, while still co-operating with other
organisations and institutions. Metadata is a way of sharing details of their
data in catalogues of geographic information, clearing houses or via vendors
of information. Metadata also gives users the information they need to process
and interpret a particular set of geospatial data.
In the mid-1990s the idea of a core set of semantics for web-based resources
was put forward for categorising the web and to enhance retrieval. This
became known as the Dublin Core Metadata Initiative (DCMI), which has
established a standard for describing web content and which is not discipline-
or language-specific. The DCMI defines a set of data elements which can be
used as containers for metadata. The metadata is embedded in the resource,
or it may be stored separately from the resource. Although developed with
web resources in mind it is widely used for other types of document,
including non-digital resources such as books and pictures. DCMI is an
ongoing initiative which continues to develop tools for using Dublin Core.
This position was questioned by Gorman (2004), who suggested that
metadata schemes such as Dublin Core are merely subsets of much more
sophisticated frameworks such as MARC (Machine Readable Cataloguing).
He suggested that without authority control and use of controlled vocab-
ularies, Dublin Core and other metadata schemes cannot achieve their aim of
improving the precision and recall from a large database (such as web
resources on the internet). His solution is that existing metadata standards
should be enriched to bring them up to the standards of cataloguing.
However, his arguments depend on a distinction being drawn between ‘full
cataloguing’ and ‘metadata’. An alternative view (and one supported in this
book) is that cataloguing produces metadata. Gorman is certainly right in
suggesting that metadata will not be particularly useful unless it is created in
line with more rigorous cataloguing approaches.
All these metadata traditions have come together as the different
communities have become aware of the others’ activities and have started to
work together. The DCMI involved the database and the LIS communities
from the beginning with the first workshop in 1995 in Dublin, Ohio, and has
gradually drawn in other groups that manage and use metadata.
Looking at existing trends, therefore, metadata is becoming more widely

recognised and it is becoming a part of the specification of IT applications
and software products. For example, ISO 15489 (ISO, 2016a), the international
standard for records management, specifies minimum metadata standards.
Library management systems, institutional repositories and enterprise
management systems handle resources that contain embedded metadata,
which they are exploiting to enhance retrieval and data exchange. As a result,
suppliers often incorporate metadata standards into their products.
This brief history of metadata demonstrates that it had several starting
points and arose independently in different quarters. In the 1990s, wider
awareness about metadata began and the work of bodies such as the Dublin
Core Metadata Initiative has done a great deal to raise the profile of metadata
and its widespread use in different communities. It has become an established
part of the information environment today. However, its history does mean
that there are distinct differences in the understanding of metadata and it is
necessary to develop some universal definitions of the term. In the time since
the publication of the previous edition of this book there have been a number
of significant developments, which are reflected in the modified chapter
structure of the book. Online social networking services have taken hold and
become a pervasive environment. This has led to unparalleled volumes of
transactional data, which is tracked and analysed to enable service providers
to sell digital advertising services. This has become a major revenue earner
for some of the largest corporations currently in existence, such as Facebook,
Alphabet and Microsoft. The data about these transactions is metadata and
this has become a tradable commodity. The concluding chapter (Chapter 14)
discusses the implications of metadata and social media.
RDA (Resource Description and Access) was in development in 2004 and
has now been adopted by major bibliographic authorities such as the Library
of Congress and the British Library, replacing AACR2. At the time of writing
BIBFRAME was due to be adopted as the replacement for MARC for encoding
bibliographic data (metadata). These developments are covered in Chapter 4
on metadata standards.
Another significant development is the establishment of services and
approaches based on the semantic web, first proposed by Tim Berners-Lee
(1998). The use of the Resource Description Framework (RDF) has facilitated
the development of linked data architecture using metadata to connect
different information resources together to create new services. Two aspects
of linked data are discussed in Chapter 12, where the practicalities of
managing metadata are covered, and in Chapter 13 where linked open data
is treated as an example of use of metadata in very large data collections.
INTRODUCTION 9
The politics of information, and in particular metadata, have become more

prominent in the intervening years between the first and second editions of
this book. A whole new chapter (Chapter 10) on information governance
covers issues of privacy, security and freedom of information. It also considers
the role of metadata in compliance with legislative requirements. The
concluding chapter (Chapter 14) also discusses some of the implications of
metadata use in the context of online advertising and in social media.
What is metadata?
Although there is an attractive simplicity in the original definition, ‘Metadata
is data about data’, it does not adequately reflect current usage, nor does it
describe the complexity of the subject.
At this stage it is worth interrogating the idea of metadata more fully. The
concept of metadata has arisen from several different intellectual traditions.
The different usages of metadata reflect the priorities of the communities that
use metadata. One could speculate about whether there is a common
understanding of what metadata is, and whether there is a definition that is
generally applicable.
Metadata was originally referred to as ‘meta-data’, which emphasises the
two word fragments that make up the term. The word fragment ‘meta’, which
comes from the Greek ‘μετα’, translates into several distinct meanings in
English. In this context it can be taken to mean a higher or superior view of
the word it prefixes. In other words, metadata is data about data or data that
describes data (or information). In current usage the ‘data’ in ‘metadata’ is
widely interpreted as information, information resource or information-
containing entity. This allows inclusion of documentary materials in different
formats and on different media.
Although metadata is widely used in the database and programming
professions, the focus in this book is on information resources managed in
the museums, libraries and archives communities. Some in the library and
information community defined metadata in terms of function or purpose.
However, in this context metadata has more wide-ranging purposes,
including retrieval and management of information resources, as we see in
an early definition:
any data that aids in the identification, description and location of networked
electronic resources. . . . Another important function provided by metadata is
control of the electronic resource, whether through ownership and provenance
metadata for validating information and tracking use; rights and permissions
metadata for controlling access; or content ratings metadata, a key component of

some Web filtering applications. (Hudgins, Agnew and Brown, 1999)
In his introduction to Metadata: a cataloger’s primer Richard Smiraglia provides

a definition that encompasses discovery and management of information
resources:
Metadata are structure, encoded data that describe the characteristics of

information-bearing entities to aid in the identification, discovery, assessment
and management of the described entities. (Smiraglia, 2005, 4)
Pomerantz (2015, 21–2) talks about metadata often describing containers for
data, such as books. He also suggests that metadata records are themselves
containers for descriptions of data and its containers and arrives at the
following definition of metadata: ‘a potentially informative object that
describes another potentially informative object’ (Pomerantz, 2015, 26). Zeng
and Qin (2015, 11) talk about metadata in the following terms: ‘metadata
encapsulate the information that describes any information-bearing entity’,
before switching their attention to bibliographic metadata and components
of metadata as described in Dublin Core. Gilliland also talks in terms of
information objects:
Perhaps a more useful, ‘big picture’ way of thinking about metadata is as the
sum total of what one can say about any information object at any level of
aggregation. In this context, an information object is anything that can be
addressed and manipulated as a discrete entity by a human being or an
information system. (Gilliland, 2016)
A further description is proposed to cover the range of situations in which

metadata is used, while still making meaningful distinctions from the wider
set of data about objects. If the object (say a packet of cereal on the super-
market shelf) is not an information resource, then data about that object is
merely data, not metadata. This is in contrast to Zeng and Qin (2015, 4), who
talk about a food label as containing metadata.
This book focuses primarily on metadata associated with documents, which
can be defined as information-containing artefacts, often held in memory
institutions such as libraries, archives and museums. Robinson (2009; 2015) has
built on the idea of the information chain, extending it beyond the original
domain of published scientific information (Duff, 1997). Buckland (1997) talks
about the document as evidence and considers how digital documents sit with
this. This thinking has also been applied to museum objects (Latham, 2012).
INTRODUCTION 11
What does metadata look like?

Some metadata is not designed for human view, because it is transient and
used for exchange of data between systems. Human-readable examples of
metadata range from html meta-tags on web pages to MARC 21 or
BIBFRAME records used for exchanging cataloguing data between library
management systems. The metadata can be expressed in a structured
language such as XML (Extensible Markup Language) or the Resource
Description Framework (RDF) and may follow guidelines or schema for
particular domains of activity.
The two examples below show metadata associated with different types of
information resource. The first is an extract taken from the British Library’s
main catalogue:
Title: Sapiens: a brief history of humankind / Yuval Noah Harari.

Author: Yuval N. Harari, author.
Subjects: Human beings — History;
Dewey: 599.909
Publication Details: London: Vintage Books, [2015?]
Language: English
Identifier: ISBN 9780099590088 (pbk)
The field names are highlighted in bold – these are equivalent to the data
elements in a metadata record. The content of each field, the metadata content,
appears alongside the field name. This same cataloguing information can be
displayed in other formats such as MARC 21.
The second example is of metadata from the home page of the Library of
Congress website, Figure 1.1 on the next page. The form displays embedded
metadata using a variety of standards. The top part of the form consists of
metadata automatically extracted from the page coding. The lower part of the
form lists metadata that the page has been tagged with according to various
metadata standards. The ‘dc:’ label refers to Dublin Core. The ‘og:’ tag refers
to Open Graph metadata.
Purposes of metadata
Metadata is something which you collect for a particular purpose, rather than
being a bunch of data you collect just because it is there or because you have
some public duty to collect (Bell, 2016). One of the main drivers for the
evolution of metadata standards is the use to which the metadata is put, its
purpose. Even within the library and information profession, a wide range
Figure 1.1 Metadata from the Library of Congress home page
of metadata purposes has been identified. Two of the most useful models
provide a basis for the purposes of metadata described in this book.
In the first model Day (2001) suggested that metadata has seven distinct
purposes. He starts with resource description – identifying and describing
the entity that the metadata is about. The second purpose is focused on
information retrieval – and in the context of web resources this is called
‘resource discovery’. This is one of the primary focuses of the Dublin Core
INTRODUCTION 13
Metadata Initiative. He recognises that metadata is used for administering

and managing resources (purpose 3) – for instance, flagging items for update
after set periods of time have elapsed. The fourth purpose, intellectual
property rights, is very important in the context of e-commerce. E-commerce
has not been listed as a purpose in its own right, possibly because Day’s
model is oriented towards web resources. Documenting software and
hardware environments, the fifth purpose provides contextual information
about a resource, but will not apply to every resource. This could be seen as
one aspect of resource description. Day’s sixth purpose, preservation
management, is a specialised form of administrative metadata and could be
incorporated into purpose 3, managing information. Finally, providing
information on context and authenticity is important in archives and records
management, where being able to demonstrate the authenticity of a record is
a part of good governance. For collection management, the provenance of
individual items may affect their value. Table 1 summarises the seven
purposes of metadata identified by Day.
Table 1.1 Day’s model of metadata purposes
1 Resource description
2 Resource discovery
3 Administration and management of resources
4 Record of intellectual property rights
5 Documenting software and hardware environments
6 Preservation management of digital resources
7 Providing information on context and authenticity
Gilliland (2016) takes a slightly different approach, although she also classifies
metadata according to purpose. The use of metadata is categorised into more
specific sub-categories. This means that a metadata scheme as well as
individual metadata elements could fall into several different categories
simultaneously. Gilliland provides some useful examples of the metadata that
falls under each type (Table 1.2). There is some common ground with Day, in
that they both identify: administration (equivalent to management and
administration); description (encompassing information retrieval or resource
discovery); and preservation as key purposes of metadata. The technical
metadata in Gilliland corresponds to ‘Documenting hardware and software
environments’ in Day. The ‘Use’ metadata could include transactional data
as would be seen in an e-commerce system or could provide an audit trail for
documents in a records management system.
Table 1.2 Different types of metadata and their functions, extracted from Gilliland (2016)
Category Definition Example
Administrative Metadata used in managing • Acquisition and appraisal information

and administering • Rights and reproduction tracking
collections and information • Documentation of legal, cultural, and
resources community-access requirements and
protocols
• Location information
• Selection criteria for digitization
• Digital repatriation documentation
Descriptive Metadata used to identify, • Metadata generated by original creator
authenticate, and describe and system
collections and related • Submission-information package
trusted information • Cataloging records
resources • Finding aids
• Version control
• Specialised indexes
• Curatorial information
• Linked relationships among resources
• Descriptions, annotations, and
emendations by creators and other users
Preservation Metadata related to the • Documentation of physical condition of
preservation management resources
of collections and • Documentation of actions taken to
information resources preserve physical and digital versions of
resources (e.g. data refreshing and
migration)
• Documentation of any changes occurring
during digitization or preservation
Technical Metadata related to how a • Hardware and software documentation
system functions or • System-generated procedural
metadata behaves information (e.g. routing and event
metadata)
• Technical digitization information (e.g.
formats, compression ratios, scaling
routines)
• Tracking of system-response times
• Authentication and security data (e.g.
encryption keys, passwords)
Use Metadata related to the • Circulation records
level and type of use of • Physical and digital exhibition records
collections and information • Use and user tracking
resources • Content re-use and multiversioning
information
• Search logs
• Rights metadata
There is a lot of common ground between these two models and although
neither of them specifically mentions ‘interoperability’ as a purpose, it is
alluded to. For instance, Day’s purpose 5 – ‘documenting software and
hardware environments’, touches on one aspect of interoperability and the
INTRODUCTION 15
Gilliland model refers to Technical metadata ‘related to how a system

functions or metadata behaves’. There is some scope for simplifying Day’s
model so that ‘Preservation management of digital resources’ (purpose 6)
becomes part of ‘Administration and management of resources’ (purpose 3),
a connection that he previously acknowledged (Day, 1999). Likewise,
‘Providing information on context and authenticity’ (purpose 7) could be
grouped with ‘Record of intellectual property rights’ (purpose 4) to become
‘Record of context, intellectual property rights and authenticity’. Gilliland’s
model could be extended by separating out the description and the
information retrieval purposes for instance.
The six-point model

This book proposes a modified, six-point model to describe the purposes of
metadata, developed from the five-point model described in the first edition.
It also separates description from retrieval as a separate, distinct purpose.
Some areas have been consolidated, such as management of resources and
preservation management (which is presented as a sub-set of management)
and rights management, which is tied in with provenance and authenticity.
This model also makes a distinction between the purposes of metadata (i.e.
the ways in which it is used) and the intrinsic properties of metadata
elements. In doing this it becomes clear that each data element can be used in
a variety of ways and fulfils more than one purpose.
The new model encompasses the purposes identified above and includes
e-commerce and information governance. The six purposes of metadata
proposed in this book are described below and provide the basis for Part II
(Chapters 5–10).
1 Resource identification and description – This is particularly important in

organisations that need to describe their information assets. For example,
under the Freedom of Information Act in the UK, public authorities have
to produce publication schemes which identify all their publications and
intended publications. In the USA, Federal agencies have to make
information available via the Government Information Locator Service
(GILS). These both depend on adequate descriptions of the data.
Information asset registers compiled by public authorities and
increasingly by the corporate sector also require descriptions of
information repositories and resources.
2 Retrieving information – In the academic sector a lot of effort has been put
into resource discovery on the internet. Aggregators and metadata
harvesting systems allow users access to material from multiple
collections. The cataloguing data usually includes a description of the

resource, controlled indexing terms and classification headings. This is a
metadata resource and may also ‘mine’ or ‘extract’ metadata directly
from target websites or electronic resources.
3 Managing information resources – The growth of electronic document and
records management (EDRM) systems and the emergence of enterprise
search systems are a consequence of operational and regulatory
requirements of large organisations. EDRM systems need access to
‘cataloguing’ information about individual records in order to manage
them effectively. Examples of metadata used in EDRM systems include:
authorship, ownership (not necessarily the same thing), provenance of the
document (for legal purposes) and dates of creation and modification.
These and other data elements provide a basis for managing the
documentation cost-effectively and consistently. Chapter 6 describes how
metadata is used to manage the retention and disposal of records.
4 Managing intellectual property rights – Metadata provides a way of
declaring the ownership of the intellectual content of an information
resource, including published documents, music, images and video. It
also provides a record of the authenticity of the document by providing
an audit trail so that, for instance, an electronic document or a digital
image will stand up in court as legally admissible evidence. One of the
preconditions for widespread acceptance of electronic documents as
original evidence is that electronic systems are becoming the preferred
medium for long-term storage.
5 Supporting e-commerce and e-government – Metadata acts as an enabler of
information and data transfer between systems, and as such is a key
component in interoperability. In order to allow software applications
that have been designed independently to pass data between them, a
common framework for describing the data being transferred is needed
so that each ‘knows’ how to handle that data in the most appropriate
manner. This may be at the level of distinguishing between different
languages, or understanding different data formats.
Interoperability is one of the enablers for e-commerce. When a piece of
data (or an aggregation of data) is passed from one system to another the
accompanying metadata (which is sometimes embedded in the digital
file) allows the new application to make sense of the data and to use it in
the appropriate fashion. For instance, in the book trade many suppliers
using different software packages need to be able to exchange data
reliably. The widely adopted ONIX standard allows different agents in
the supply chain from author to reader to exchange data without the
need to integrate their systems.
INTRODUCTION 17
6 Information governance – Information governance is now an established

area of metadata application. It can be used to provide an audit trail for
data collections, for instance. This allows compliance managers to
demonstrate that they are handling data in an appropriate fashion. For
example, sensitive personal data needs to be kept securely, with access
limited to specified individuals. Freedom of information legislation, on
the other hand, may require a retention schedule and publication scheme
to be associated with specific information resources. Some metadata
standards have data elements specifically geared to recording an audit
trail associated with a document.
Multiple purposes
Metadata can be used within one application for several different purposes.
The model developed here helps in the analysis of metadata applications and
the understanding of its characteristics in different situations.
Why is metadata important?

A more comprehensive understanding of metadata can be developed from
studying the above examples. The development of cataloguing over more
than two millennia has provided a set of tools for describing published
information. This has been drawn on by the web community.
Correspondingly, the growth of the internet has focused public attention on
the importance of information retrieval and management and has stimulated
the development of tools to improve retrieval performance. Having a clear
understanding of what metadata is and how it is managed provides a means
of handling information resources more effectively.
Organisation of the book

This book is arranged in three sections. Part I (Chapters 1–4) deals with the
fundamental concepts of metadata and can be seen as an introduction to the
subject. It is pitched at the community of information professionals and users
such as academics that are interested in metadata for managing and retrieving
documentary information or information resources. The book uses the terms
‘document’ in the widest sense as a vehicle for information communications
(Robinson, 2009).
Part II (Chapters 5–10) considers the purposes of metadata from
identification of information resources to retrieval, and onwards to e-
commerce applications and information governance. This builds on the five
purposes identified in the first edition and has been extended and modified
to reflect the full range of uses of metadata in the 14 years that have since
passed.
Part III (Chapters 11–14) is about the management of metadata and starts
with well established methods of managing standards, schemas and metadata
quality. It then considers recent developments in taxonomies, encoding
schemes and ontologies and the role that these play in structuring knowledge.
It moves on to big data and the challenges faced by those wishing to exploit
very large data collections. It then considers the starting point of this book,
politics. What are the implications for privacy and national security? The final
chapter also considers the future of metadata – from the empowerment of
users through to professional development – and considers who will be
responsible for managing metadata in the future.
Throughout this book ‘metadata’ is used as a singular collective noun. The
word ‘data’ is used as a mass noun and is treated as a collective singular noun
in accordance with most common current usage (Rosenberg, 2013, 18–19).
This ties in with the gradual disappearance of the word ‘datum’. Even Steven
Pinker, one of the foremost thinkers and writers about linguistics
acknowledges this, although he makes clear his own preferences:
I like to use data as a plural of datum, but I’m in a fussy minority even among
scientists. Data is rarely used as a plural today, just as candelabra and agenda
long ago ceased to be plurals. But I still like it. (Pinker, 2015, 271)
CHAPTER 2
Defining, describing and expressing metadata
Overview
This chapter describes some of the concepts associated with metadata. It considers ways
in which metadata can be expressed and focuses on document mark-up languages. It
then considers schemas as one method of defining metadata standards and data
elements. Databases of metadata are described as an alternative to embedded
metadata. The last section of the chapter shows some examples of how metadata is used
in different contexts such as document creation, records management, library catalogues,
digital repositories and image collections.
Defining metadata
Metadata is used in catalogues of digital and physical information resources.
The requirements for books in a library catalogue might be very different
from the metadata embedded in a web page, but the general concepts of
metadata apply to both. Its use for digital and printed resources provides
some helpful examples. Document mark-up languages such as SGML and
XML are widely used to express metadata standards.
Document mark-up
The development of mark-up languages is an excellent example of the way
in which metadata can be applied to and expressed in documents. Electronic
documents are one of the most common forms of digital object to which
metadata is applied, and range from web pages through to electronic records,
and may incorporate text, images and interactive material.
Mark-up languages were initially developed to describe the layout and

presentation of documents. They enabled organisations to manage large
numbers of documents that needed to be presented in different formats.
Mark-up languages also provide a means of defining metadata standards.
Mark-up languages, which arose from text processing, are defined as:
‘computer systems that can automate parts of the document creation and
publishing process’ (Goldfarb and Prescod, 2001). Mark-up languages
containing a combination of text and formatting instructions include:
• HTML
• XML
• TEI
• LaTeX.
Figure 2.1 shows: raw text (the data), the text with formatting instructions
(the mark-up) and the text as it would appear to the reader (the rendition).
This is an example This is an example of

of marked-up text marked-up text that This is an
that shows large shows <l>large</l> example of
and small text as and <s>small</s>text marked-up text
well as bold and as well as that shows
italics <b>bold</b> and large and small
<i>italics</i> text as well as
bold and italics
data mark-up rendition

(raw text) (text with mark-up (text as it appears
instructions) to the reader)
Figure 2.1 Example of marked-up text
Standard Generalized Markup Language (SGML)

Standard Generalized Markup Language (SGML) is used as the basis for
describing many web pages and for marking up metadata. Generalised
document mark-up originated in the late 1960s from the work of three IBM
researchers, Goldfarb, Mosher and Lorie, whose initial letters make up the
‘GML’ in SGML (Goldfarb, 1990). They determined that a mark-up language
would need three attributes:
• Common data representation – so that different systems and applications

are able to process text in the same representation
DEFINING, DESCRIBING AND EXPRESSING METADATA 21
• Mark-up should be extensible – so that it can support all the different

types of information that must be exchanged. There is potentially an
infinite variety of document types that can be generated
• Document types need rules – formal rules for documents of a particular
type, which can be used to test their conformance to the type and
therefore how they are processed.
These attributes provide a framework for representing metadata. A common

representation is needed so that metadata elements are clearly identifiable
and can be processed appropriately. The extensibility of mark-up languages
allows considerable flexibility in creating metadata tags. Document types are
used to describe the ‘rules’ for metadata schemas, so that there is consistency
in their expression. The development of a generalised mark-up language
ensured that documents could be handled in a variety of environments.
Rather than focusing on formatting instructions, a generalised mark-up
language tags different data types. A stylesheet translates generalised mark-
up into formatting instructions. For instance, it can instruct a system to make
section headings in bold text and quotations in italics. Different stylesheets
can be applied to the same marked-up text. This means that the same text can
be presented in different ways, for instance as a printed publication, or
displayed as a web page viewed with a browser (Figure 2.2).
This is an
example of
marked-up text
that shows
large and small
This is an example text as well as
of marked-up text Stylesheet 1 bold and italics
that shows
<l>large</l> and
<s>small</s>text as
well as <b>bold</b> This is an example of
and <i>italics</i> marked-up text that
shows large and small
text as well as bold
and italics
Stylesheet 2
Marked-up text Formatting Text as it is displayed

instructions to readers
Figure 2.2 Rendered text

SGML was for a long time an international standard, ISO 8879 (ISO, 1986).
Although now withdrawn, it is still the basis for other mark-up languages.
Hypertext Markup Language (HTML) is an application of SGML. HTML is
used to encode the content of web pages and is widely used to describe web
pages, including the metadata embedded in them. HTML5 recognises
metadata content as a specific category of HTML content:
Metadata content is content that sets up the presentation or behavior of the rest of
the content, or that sets up the relationship of the document with other
documents, or that conveys other ‘out of band’ information.
(W3C, 2014b)
The head of the document holds the metadata, including the title and other
metadata content held in the data element ‘meta’.
TEI is a specialist mark-up language widely used in the digital humanities
(TEI Consortium, 2016). The TEI header is where metadata is normally
embedded. However, marking up documents in this way allows other
characteristics of a document to be identified and retrieved or processed. The
Title Statement includes title, author, and funder information. Other
bibliographic information includes edition and publication details. Although
TEI is not a cataloguing standard, its structure facilitates identification and
use of structured metadata.
LaTeX is another specialist mark-up language, developed for scientific and
mathematical publications. Different templates can be applied to a marked-
up document to format it to conform with a variety of academic publications.
The current version is LaTeX2e. LaTeX3 was still in development at the time
of writing (LaTeX3 Project Team, 2001).
XML (Extensible Markup Language)

Extensible Markup Language (XML) is a subset of SGML. It offers the ability
to represent data in a simple, flexible, human-readable form. As an open
standard, XML is not controlled by one vendor or one country. The XML
specifications are published by the World Wide Web Consortium (W3C), an
international co-operative venture (W3C, 2016). XML can be used as a basis
for exchange of data or documents between people, computers and
applications. It goes further than HTML because it provides a way of
expressing a semantic context for data, as well as dealing with the syntax. It
is the semantic component which gives XML the ability to exchange data in
a meaningful way and this is one of the reasons for its widespread uptake.
XML handles characters, which are made of character data (the text or data
content) and mark-up, which encodes the logical structure and other
attributes of the data. Documents are organised into elements which break
the document down into units of meaning, purpose or layout. The elements
correspond to fields in a database, as will be seen in later examples in this
chapter. XML documents can also use entities, which may refer to an external
document or a dynamic database record, or can be used to label a defined
piece of text for re-use within the document.
Document type definitions (DTDs)

Cole and Han (2013) provide an excellent description of XML in the context
of cataloguing and metadata use. They describe in a step-wise process the
way in which details about the semantics of a document can be embedded in
an XML document. This depends on following a syntax (or grammar)
specifying the way in which information about a document is expressed. If
this syntactic information (rules for the organisation of content within a
document) is held in a separate document, a DTD, it can then be referred to
by multiple documents. This makes the management of the syntax rules (or
grammar) much simpler and in theory any system that renders XML
documents should follow the reference to the DTD in order to understand
the way in which the fields are formed.
A class of similar documents can be called a ‘document type’. A Document
Type Definition (DTD) is a set of rules for using XML to represent documents
of a particular type. DTDs provide one form of metadata expression in mark-
up languages, as they refer to the vocabulary and the rules used to describe
metadata.
The DTD defines the elements (or fields) of a document. This means that
similar documents can be defined by the same DTD. For instance, a memo
might have the following elements:
To: (the addressee)

From: (the author)
Date: (date on which the memo was sent)
Subject: (what the memo is about)
Body: (the main text of the memo)
The DTD for a memo can be used to test the ‘validity’ of the document. In
other words does a document purporting to be a memo have the right
elements appearing in the right order? If it does, the DTD provides the means
for the memo to be expressed in a variety of formats determined by the
appropriate stylesheet. In this example, the ‘Memo’ DTD might have separate
stylesheets for printed-out memos, screen displays, and e-mail versions.

Carrying on with the memo example, the elements are delimited by tags. The
‘To’ element could be expressed by the following tags:
<!ELEMENT To>Jane Williams</To>
The element may have attributes associated with it – in terms of the encoding
system used for instance, or the type of data that appears in that element. For
example the ‘To’ element could be defined by the following statement:
<!ELEMENT To (#PCDATA)>
This indicates that the ‘To’ data element consists of Parsed (parsable)
Character data (#PCDATA).
XML schemas
An alternative way of defining metadata is to use XML schemas. They offer
greater flexibility than DTDs and are widely used for expressing metadata
standards. Schemas are XML languages used for defining similar types of
document in terms of their structure, content and meaning. The W3C website
defines them in the following terms:
XML Schemas express shared vocabularies and allow machines to carry out rules
made by people. They provide a means for defining the structure, content and
semantics of XML documents. (Sperberg-McQueen and Thompson, 2014)
They are described using XSDL (XML Schema Definition Language). The
following extracts are from an example of an XML schema that defines simple
Dublin Core metadata elements (Cole et al., 2008).
The start of the schema contains declarations about the nature of the
schema, including two namespace references ‘xmlns’.
<?xml version=‘1.0’ encoding=‘UTF-8’?>

<xs:schema xmlns:xs=‘www.w3.org/2001/XMLSchema’
xmlns=‘http://purl.org/dc/elements/1.1/’
targetNamespace=‘http://purl.org/dc/elements/1.1/’
elementFormDefault=‘qualified’
attributeFormDefault=‘unqualified’>
This is followed by annotations from the authors about the background to the
schema and then a namespace reference to the standard for XML.
<xs:annotation>
<xs:documentation xml:lang=‘en’>
DCMES 1.1 XML Schema
XML Schema for http://purl.org/dc/elements/1.1/namespace
Created 2008-02-11
Created by
Tim Cole ([email protected])
Tom Habing ([email protected])
Jane Hunter ([email protected])
Pete Johnston ([email protected]),
Carl Lagoze ([email protected])
This schema declares XML elements for the 15 DC elements
from the http://purl.org/dc/elements/1.1/namespace.
It defines a complexType SimpleLiteral which permits mixed
content
and makes the xml:lang attribute available. It disallows child
elements by use of minOcccurs/maxOccurs.
However, this complexType does permit the derivation of other
complexTypes which would permit child elements.
All elements are declared as substitutable for the abstract
element
any, which means that the default type for all elements is
dc:SimpleLiteral.
</xs:documentation>
</xs:annotation>
Namespace declarations can also be used to link to a metadata standard or

encoding scheme at the start of a record. The main body of the schema defines
the 15 data elements in simple Dublin Core.
<xs:element name=‘any’ type=‘SimpleLiteral’ abstract=‘true’/>

<xs:element name=‘title’ substitutionGroup=‘any’/>
<xs:element name=‘creator’ substitutionGroup=‘any’/>
<xs:element name=‘subject’ substitutionGroup=‘any’/>
<xs:element name=‘description’ substitutionGroup=‘any’/>
<xs:element name=‘publisher’ substitutionGroup=‘any’/>
<xs:element name=‘contributor’ substitutionGroup=‘any’/>
<xs:element name=‘date’ substitutionGroup=‘any’/>
<xs:element name=‘type’ substitutionGroup=‘any’/>
<xs:element name=‘format’ substitutionGroup=‘any’/>
<xs:element name=‘identifier’ substitutionGroup=‘any’/>
<xs:element name=‘source’ substitutionGroup=‘any’/>
<xs:element name=‘language’ substitutionGroup=‘any’/>
<xs:element name=‘relation’ substitutionGroup=‘any’/>
<xs:element name=‘coverage’ substitutionGroup=‘any’/>
<xs:element name=‘rights’ substitutionGroup=‘any’/>
Schemas are commonly associated with databases, where each data element
corresponds to a field in a database. As with databases, the schema can be set
up to provide semantic and syntactic checks on data. In other words, checks
on the meaning and grammar of an expression can be made. Syntactic checks,
for example, can be applied to the data to ensure that it is of the appropriate
type and is expressed in a format that can be processed by the database
software. For example, dates can be defined using international standard ISO
8601:2004 to get over the problem of differing American and British date order
(ISO, 2004c). For instance, ‘10/12/17’ means ‘10th December 2017’ in Britain
and ‘October 12th 2017’ in the USA. Schemas can also apply semantic checks
to ensure that business rules are followed by requiring the value of an element
(the field content) to fall within a specified range. For instance, the value of
the month element in the data should be between 1 and 12. The
www.schema.org website offers a resource for sharing schemas of this type
and this is described in more detail in Chapter 12.
Namespace
Namespace is used to locate definitions for metadata schema from the Internet. This
ensures greater consistency of terminology used to define metadata elements and
provides a way of sharing elements. In the Dublin Core example the namespace that
provides the original reference to Dublin core elements is as follows:
xmlns=‘http://purl.org/dc/elements/1.1/’
A formal definition is (Bray et al., 2009):
An XML namespace is identified by a URI reference [RFC3986]; element and

attribute names may be placed in an XML namespace using the mechanisms
described in this specification.
Databases of metadata
The previous section about the mark-up of documents focused particularly
on embedded metadata. For example, a web resource may have metadata
tags and content embedded in the resource. Electronic documents and other
digital materials often have embedded metadata, allowing other applications
and systems to effectively process them. However, this is not the only way of
handling metadata. In many systems the metadata may be held separately in a
database.
Databases of metadata may be generated at the point of creation of
documents by Enterprise Content Management (ECM) systems, for instance.
ECM systems store the metadata about documents in a central database and
use this data to manage and handle the documents. This allows documents
to be brought forward for review, the workflow to be managed, and access
to be controlled. Institutional repositories and library management systems
operate in a similar fashion, working with central collections of metadata, the
library catalogue or repository database.
Examples of metadata in use

Data can be catalogued in a variety of ways, which are demonstrated by some
of the examples described here. Common applications such as word processors,
ECM systems, library catalogues and directories of digital repositories all make
extensive use of metadata. Some of these application areas are described more
fully in the chapters about the specific purposes of metadata.
Word-processed documents
Applications that are used for preparing documents, such as word-processing
packages, automatically generate metadata when a document is saved for the
first time. In some cases systems can be configured to prompt the author for
metadata when a new document is saved. Metadata associated with the user
such as ‘Author’ and ‘Company’ may be automatically generated. This can be
edited and additional metadata can be added manually. A controlled
vocabulary can be a useful way of ensuring consistent retrieval of documents.
For instance, keywords selected from a thesaurus can be added as metadata
to enrich the subject description of the document.
The screenshot in Figure 2.3 on the next page shows a typical metadata
screen associated with a word-processed document. In Microsoft Word there
are additional tabs for: ‘Statistics’, and ‘Contents’, which display metadata
such as document size, time spent editing it, the session number, and the
number of words. The final tab ‘Custom’ allows for additional optional data
associated with document and records management.
Electronic records management

The word-processing example has shown metadata designed for human use
and often requiring human intervention. However, metadata is increasingly
used by computer applications without direct human intervention. ECM and
EDRM systems make use of metadata associated with documents to manage
them effectively. While many of these data elements can be examined by
human beings, they are used by the software to process records during their
lifecycles.
Figure 2.3 Word document metadata
An example of this is the retention period of a record. If a record is assigned

a specific category according to a file plan (usually a business classification
scheme applied to an organisation’s records), there will be an associated
retention period. Typically each category in the file plan will have a set
retention period. For example, in a recruitment exercise, interview records
may be kept for six months from the interview date before disposal (unless
the candidate is successfully recruited, in which case they become part of the
employee record with its own retention rules). Invoices may be kept for six
years, on the other hand, to comply with company legislation in some

countries. The file category assigned to a document or record is encoded as a
metadata element that the EDRM system uses to identify records that are due
for review or disposal at the end of each retention period. In order to do this
the system will have to call on another metadata element containing
information on the date created. The system can then generate a disposal list
for review by a records manager, or administrator.
Library catalogues
Metadata is particularly useful for large collections of documents or other
materials, where it can be used for managing the resource and for finding
specific items. A catalogue becomes essential for retrieval when there are
more than a few hundred items in a collection. The arrangement of books in
a subject classification is not always sufficient for good subject retrieval. If a
book is about more than one subject, it can only be held in one place
physically. It may also be out on loan, which means that shelf browsing would
not identify it. Early examples of library metadata were held on catalogue
cards. In the 1960s electronic catalogues began to appear and are routinely
used in most libraries today. Library users or patrons can find books by
searching the catalogue by a variety of criteria such as author name, words
in the title, classification code (which determines the arrangement on the
shelves) and keyword (subject).
Figure 2.4 on the following page shows the results of a search on the subject
‘Sherlock Holmes’ in the catalogue from the City of Westminster Libraries,
London. In this example the metadata is used to store comparable data about
individual items in the collection. This allows users to search consistently
across the whole collection. Other metadata associated with items, such as
location (library branch), author and format are all metadata elements that
can be used to refine the results of the search. Within each item, other
metadata such as title, publication date, abstract and availability are
displayed.
The detailed record shows additional metadata elements, including ISBN,
subject terms, physical description and genre. An even more detailed system
record such as that shown in Figure 2.5 will be available to library staff, which
will contain administrative data such as date of acquisition, accession number
and the status of the item in collection management processes such as
labelling, repair and withdrawal.
Figure 2.4 Westminster Libraries – catalogue search © 2017, Westminster

City Council and Sirsi Corporation
Figure 2.5 Westminster Libraries catalogue record © 2017, Westminster

City Council and Sirsi Corporation
Searching a group of catalogues

When searching across a group of library catalogues, metadata allows users
to access several differently structured catalogues at once. Common
cataloguing rules such as RDA (and previously AACR2) help to ensure a
degree of consistency of each data element across systems. For instance, the
author name might take the form Surname, First Name.
The WorldCat union catalogue provided by OCLC is the largest in the
world with (at the time of writing) over 300 million catalogue records pointing
to over 2 billion individual items in 72,000 libraries in 140 countries. In the
example in Figure 2.6, a search on Francis Kéré, the architect from Burkina
Faso, yields 317 items where his name is mentioned.
Figure 2.6 WorldCat search © 2017, OCLC, Inc.
Selecting one item from the results list provides a detailed catalogue entry for
that item. Figure 2.7 on the next page is the WorldCat entry with
supplementary information about holdings in participating libraries.
Figure 2.7 WorldCat detailed reclord © 2017, OCLC, Inc.
Digital repository search

Another kind of search operates from a central database of details about
digital repositories. In Figure 2.8 a search of OpenDOAR (the directory of
open access repositories) for Brazil lists 86 repositories. It is possible to narrow
down the search by a number of different criteria which can be selected from
drop-down lists, such as subject area, language, content type, repository type
or by free text.
Figure 2.8 OpenDOAR search of repositories

A closer look at individual records (Figure 2.9) shows the metadata about the
repository (repositories, being information resources themselves, are
described by metadata).
Figure 2.9 Detailed OpenDOAR record
Image repositories
Image repositories such as iStockPhoto, Getty Images and Flickr use specialist
metadata as well as keywords to help people to retrieve images (Getty Images,
2017; iStockphoto LP, 2017; Yahoo! Inc., 2017). So, for instance, it is possible
to search Getty Images by the following criteria: image type, orientation,
number of people, colour, image size, age (of people in the image), people
composition, image style, ethnicity, photographers, and royalty-free
collections.
Conclusion
These examples of metadata are based on the principle that metadata may be
embedded in a digital object or held separately from the resource that it
describes. Mark-up languages such as XML provide a way of handling and
exchanging metadata. They also provide a means of describing metadata
standards.
CHAPTER 3
Data modelling
Overview
Metadata models help to give an understanding of the development of metadata
standards. The chapter starts with an overview of data modelling and its relationship to
metadata. It defines some of the terminology used to describe modelling languages,
using the Unified Modelling Language (UML) as an example. Systems such as RDF and
the ABC Ontology are discussed before considering domain-specific modelling
frameworks such as the Library Reference Model (LRM), indecs for the book trade and
OAIS for online exchange of information. Van Hooland and Verborgh (2014) talk about
four types of modelling: tabular, relational, meta mark-up and RDF. This chapter focuses
on the last two types.
Metadata models
A metadata standard is a type of data model which provides a way of
conceptualising the characteristics of an information resource. A data model
may have its own syntax (grammar) and semantics (meaning) and may be
expressible in a mark-up language such as XML (W3C, 2016). One of the
interesting aspects of the development of metadata standards has been the
convergence of different communities of interest. People have recognised the
benefits of working within common frameworks. In order to do so they have
adopted common languages for describing the data that they handle.
Languages such as XML and RDFa have played an important role in equip-
ping these communities with a set of tools to describe data and relationships
between data elements (Herman et al., 2015).
Unified Modelling Language (UML)

UML provides a framework for describing data and for data modelling. Hay
(2006) talks about ‘semantic data constructs’ when discussing metadata. He
introduces the idea of different levels of data, starting with real world things.
These might be people or books or items in a store. The next level up is ‘data
about real world things’. Typically this would consist of data in databases,
including HR systems (people), a library catalogue (books) or a stock control
system (items in a store). At the next level up is data about a database or
‘metadata’. This in effect is a data model because it describes the structure of
the data. An example from the database domain would be a data dictionary,
which is a list of the fields in a database and their characteristics.
UML originates from the Object Management Group and is an ongoing
initiative with organisational members (OMG, 2015). ISO 19505-1, the
equivalent international standard, states: ‘The Unified Modelling Language
is a visual language for specifying, constructing, and documenting the
artifacts of systems.’ (ISO, 2012a, 8).
Resource Description Framework (RDF)

The Resource Description Framework (RDF) is a system for modelling data
and is a way of expressing metadata about an information resource or
information object (Schreiber and Raimond, 2014). RDFa is a language based
on RDF for representing information about resources on the internet (Herman
et al., 2015). It allows for exchange of this information on the web and for
processing by applications. RDF was one of the first tools developed for
modelling and describing web resources. It goes beyond metadata description
by providing a model for the relationships between different metadata
elements. It bridges the divide between human-generated and machine-
generated (and machine-processed) metadata. RDF works with different
types of object or data entities and defines the relationship between them.
RDF can be expressed in mark-up languages such as HTML and XML. Its
purpose is to enable the encoding, exchange and re-use of metadata
definitions or schema. The system is flexible, allowing each resource
description community to define its own metadata elements. It also allows
those communities to tap into existing schemas and to re-use elements that
may be relevant. The namespace convention ensures that there is a unique
reference back to the original definition. This system exploits the power and
range of the internet and avoids the need for a central register or repository
of data elements. As an object-oriented system RDF is based on three object
types:
DATA MODELLING 37
1 Resources – anything being described by RDF expressions. This could be

a web page or a printed book, for instance.
2 Properties – an attribute or characteristic of the resource being described.
For instance, ‘Creator’ can be applied to a web page, or ‘Author’ is a
property of a book. The schema specification will describe how that
property is expressed. For instance, cataloguing rules may require
authors to take the form: Surname, Initials. For example, the author Jane
Smith would be expressed as ‘Smith, J.’.
3 Statements – a statement applies to a specific resource and includes a
subject (the resource), the predicate (the property) and object (the value
of the property).
The statement syntax: subject – predicate – object is known as a triple or 3-

tuple (Figure 3.1). The following statement describes the author of a book and
can be represented by the triple:
The book, The Hound of the Baskervilles (subject) has creator (predicate)
Arthur Conan Doyle (object).
Creator
The Hound of the Arthur Conan Doyle
Baskervilles
Figure 3.1 An RDF triple
This structure is recursive, so that the subject of a triple can be another triple.
In other words there can be metadata about the metadata. It is also possible to
chain statements to produce more detailed metadata records, as illustrated in
Figure 3.2. The author name is represented by a Uniform Resource Identifier
(URI) or an Internationalized Resource Identifier (IRI) with properties of ‘Name’
The Hound of the Creator

http://viaf.org/viaf/65283845
Baskervilles
Name Lifespan
Arthur Conan Doyle 1859–1930
Figure 3.2 More complex RDF triple

and ‘Lifespan’ associated with it. In the ‘node and arc’ diagram below the object
of the statement is itself a statement. ‘The Hound of the Baskervilles has creator
. . .’ leads to the statement ‘URI http://viaf.org/viaf/65283845 has name Arthur
Conan Doyle and has lifespan 1859–1930’.
The mandated use of URIs in RDF makes it a powerful tool for creating
linked data. Examples of its use in open data initiatives can be found in
Chapter 13. The author statement in Figure 3.3 can be expressed in the
following terms:
The Hound of the Baskervilles has author Arthur Conan Doyle

http://viaf.org/viaf/181625655 has dc:creator http://viaf.org/viaf/65283845
http://viaf.org/viaf/181625655 http://purl.org/dc/terms/creator
http://viaf.org/viaf/65283845
http://purl.org/dc/
terms/creator
http://viaf.org/viaf/181625655 http://viaf.org/viaf/65283845
Figure 3.3 A triple expressed as linked data
Looking at the construction of an RDF statement expressed in XML syntax

helps to show the way in which RDF works. In this example the RDF
container is surrounded by a pair of tags, opening with:
<rdf:RDF>
and closing with:

</rdf:RDF>
The opening RDF statement includes the RDF namespace declaration, which
refers to the specific URI. This allows multiple and consistent use of XML
resources, because different documents can refer to the same namespace. It
also ensures that an application can recognise and use the appropriate version
of RDF to interpret the statements that follow.
<rdf:RDF xmlns: rdf=www.w3.org/1999/02/22-rdf-syntax-ns#

xmlns:dc=‘http://purl.org/dc/elements/1.1/’>
Expressed with URIs to allow for linked data activity:

DATA MODELLING 39
<rdf:Description rdf:about=http://viaf.org/viaf/181625655
<dc:creator><rdf:Description
rdf:about=‘http://viaf.org/viaf/65283845’></rdf:Description></dc:creator>
</rdf:Description>
In summary, RDF is a modelling system widely used to analyse web

resources (objects) and the relationships between different entities or data
elements associated with the resources. Its expressiveness and recursive
nature allows for complex entities (such as detailed bibliographic citations)
to be represented. It also provides a language for exchange of metadata
between systems so that new services can be developed and presented to
users.
Dublin Core
The Dublin Core metadata standard, which is described in Chapter 4, is a
widely used metadata standard for describing online resources (DCMI, 2012).
It is underpinned by a data model which can be represented in UML (Powell
et al., 2007). Figure 3.4 shows the DCMI Resource Model, showing the
relationship between a resource description and establishing the fact that each
property-value pair contains one property and one value. For example, the
property ‘dc:creator’ might have the value ‘Arthur Conan Doyle’. A resource
resource
literal value
1 value
non-literal
property- value
described
resource value pair
1
Key
property
label relationship defined by label
‘is’ or ‘is a’
‘contains n’ or ‘has n’
Figure 3.4 DCMI resource model (DCMI, 2012, licensed under CC BY 4.0)
may be described using multiple property-value pairs, such as creator, title,

identifiers (ISBNs, DOIs etc.).
The Library Reference Model (LRM) and the development of RDA

RDA (Resource Description and Access), the cataloguing standard that
replaces AACR2, is based on the Library Reference Model, which superseded
the Functional Requirements for Bibliographic Records (FRBR), and the
Functional Requirements for Authority Data (FRAD) (Joint Steering
Committee for Development of RDA, 2014; IFLA, 1998; IFLA, 2013; Riva, Le
Boeuf and Žumer, 2017). The Functional Requirements for Subject Authority
Data (FRSAD) was introduced to RDA in 2015 and has also been incorporated
into LRM (Galeffi et al., 2016).
LRM is based on entity-relationship modelling and covers bibliographic
data. It is designed to support five generic user tasks summarised below
(Riva, Le Boeuf and Žumer, 2017,13):
Find – To bring together information about one or more resources of

interest by searching on any relevant criteria.
Identify – To clearly understand the nature of the resources found and to
distinguish between similar resources.
Select – To determine the suitability of the resources found, and to be
enabled to either accept or reject specific resources.
Obtain – To access the content of the resource.
Explore – To discover resources using the relationships between them and
thus place the resources in a context.
LRM defines relationships between entities which have particular attributes.

Entities are ‘key objects of interest to users of library information systems’
(Riva, Le Boeuf and Žumer, 2017). The top level entity is res (‘thing’ in Latin).
The eight entities at the next level are: work, expression, manifestation, item,
agent, nomen (‘name’ in Latin), place, and time-span. The relationships between
four of these entities are core to RDA and are illustrated in Figure 3.5 opposite:
• Work – a distinct intellectual or artistic creation

• Expression – intellectual or artistic realisation of a work
• Manifestation – physical embodiment of an expression of a work
• Item – single exemplar of a manifestation.
For example, the work Sapiens: a brief history of humankind by Yuval Noah
Harari, published in the UK in 2014, was originally published in Hebrew in
DATA MODELLING 41
WORK
is realized through
realizes
EXPRESSION
is embodied in
embodies
MANIFESTATION
is exemplified by
exemplifies
ITEM
Figure 3.5 Relationships between Work, Expression, Manifestation and Item (based on Riva,
Le Boeuf and Žumer, 2017)
Israel in 2011 and has been widely translated. Each translation could be seen
as a work or as an expression of the original Hebrew publication. If the
English version is seen as a work (with a relationship to the original Hebrew
work) an expression might be the edition published by Harvill Secker in
London in 2014. The hardback edition with ISBN 9781846558238 is a manifest-
ation of the English-language version. An example of that manifestation is
the item, which is the copy of the hardback English-language edition that is
on my bookshelf at home.
Figure 3.6 overleaf shows in general terms the relationship between a work,
expression, manifestation or item and the responsible agent (which could be
a person or a corporate body). The double-headed arrows indicate that there
may be multiple instances of a relationship between entities. For instance a
work is created by a person (or persons). The reverse relationship is that a
person creates a work (or works).
In the example title Sapiens, the work was created by Yuval Noah Harari.
The English edition in hardback was created, manufactured and distributed
by publisher Harvill Secker. The copy on my bookshelf is owned by me.
WORK
was created by
created
EXPRESSION
was created by
created
is distributed by AGENT
was created by
was manufactured by
manufactured
created
MANIFESTATION distributed
was modified by
is owned by
owns
modifies
ITEM
Figure 3.6 LRM agent relationships
LRM provides a way of analysing intellectual works such as published

books and articles. This has had a profound effect on cataloguing practice, so
that for instance, the common bibliographic elements of different expressions
of a work are catalogued only once at work level and the same basic record
is refined with additional fields that apply at expression, manifestation and
item levels. However, alternative bibliographic models such as BIBFRAME
reject the expression and manifestation entities (Baker, Coyle and Petiya, 2014;
Library of Congress, 2017a). The complexities of converting from AACR2 to
RDA should not be underestimated. Apart from identifying equivalent
records using AACR2, there is the challenge of reconciling variations in the
higher-level metadata. For instance, the work-level metadata of two
manifestations may not match exactly. Additionally there may be some
ambiguity between expression and manifestation.
ABC Ontology and the semantic web

The ABC Ontology is ‘a basic model and ontology that provides the notional
basis for developing domain, role or community specific ontologies’ (Lagoze
DATA MODELLING 43
and Hunter, 2002). The model is intended to provide a basis for analysing
existing metadata ontologies, to give communities the tools to develop their
own ontologies and to provide a mechanism for mapping between metadata
ontologies.The ABC Ontology was developed to facilitate interoperability
between metadata ontologies from different domains. Its target is to ‘model
physical, digital and analogue objects held in libraries, archives and museums
and on the Internet’ (Lagoze and Hunter, 2002). This includes books, museum
objects, digital images, sound recordings and multimedia resources. It can
model abstract concepts such as intellectual content and time-based events
such as a performance or a lifecycle event that happens to an object such as
publication of a book. The model is based on a primitive category ‘Entity’,
with three categories at the next level: Temporality, Actuality and Abstraction.
The data elements used fall into four main categories, as shown here:
• ENTITY
– Time
– Place
• TEMPORALITY
– Situation
– Event
– Action
• ACTUALITY
– Artifact
– Agent
• ABSTRACTION
– Work
Each category has subcategories that allow for more precise descriptions of
the models. These in turn can be broken down into subclasses specific to a
particular domain, such as libraries, museums or web resources. The ABC
Ontology allows for modelling of time-dependent relationships, which are
particularly important in museums and archives (where the provenance of
an item is key to its integrity), rights management (where it is important to
track who has used a work under what conditions and when) and for events
such as a musical performance. Figure 3.7 on page 44 is a simple
representation of a publication using the ABC Ontology. This is a simplified
representation of part of the publishing process. The work Omeros by the late
Nobel Laureate Derek Walcott is expressed as a book. The book is manifest
as the edition published by Faber & Faber Ltd (Walcott, 1990). A more
complete representation of this would indicate the place of publication and
co-publishing details.
“Derek Walcott” “Faber & Faber Limited”

value value
AG0 AG1
participant
1990 participant
author publisher
type atTime type hasParticipant
type authoring 1990
atTime
precedes follows type

EV0 ST0 EV1 “publishing”
creates
MN0
format
book
hasRealisation
title
WK0 “Omeros”
Figure 3.7 Publication details using the ABC Ontology
One consequence of the sophistication of the ABC Ontology, which allows

for complex modelling of entities and the relationships between them, is the
considerable effort required to analyse elements and relationships. It therefore
only becomes worthwhile if the resulting benefits are sufficiently large, as in
e-commerce applications.
Indecs – Modelling book trade data

The indecs metadata framework was developed to provide a basis for
interoperability in e-commerce: ‘In the indecs framework, interoperability
means enabling information that originates in one context to be used in another
in ways that are as highly automated as possible’ (Rust and Bide, 2000, 6).
Indecs is used to identify and describe items from different data sources.
Its use has been incorporated into ONIX, used by publishers for exchange of
metadata about their products and for Digital Object Identifiers (DOIs) (see
Chapter 5). Indecs centres on four axioms about e-commerce:
• Axiom 1: Metadata is critical – In order to trade electronically you need

information about who is trading, what is being traded and the nature of
DATA MODELLING 45
the transaction. These are all metadata. A common understanding of the

metadata elements is necessary for a successful transaction to take place.
• Axiom 2: Stuff is complex – An item such as a recording may have many
separate tracks that each carries its own rights. For instance, a recording
of a piece of music on a CD may have rights associated with the
composer, the publisher, the conductor, the performers, the recording
studio, the text used for the sleeve notes and any illustrations that are
used for the cover.
• Axiom 3: Metadata is modular – Each entity must have its own metadata,
even if they are part of a larger item, if the rights associated with them are
to be protected. The modules are linked together as a metadata network.
• Axiom 4: Transactions need automation – For e-commerce to work, it is
important that local data standards and systems are standardised. This
opens the way for automation of rights transactions and makes it
possible to handle the very large volume of requests that would come in
to a rights holder.
The indecs framework has defined metadata elements, each of which has an
indecs identifier or iid. The indecs framework can be used to model the
relationships between entities. Indecs is based on the premises that: ‘People
make stuff’, ‘People use stuff’ and ‘People do deals about stuff’ (Figure 3.8).
make
people stuff
used by
do about
deals
Figure 3.8 Indecs model
More complex models can be developed to reflect detailed transactions and

to represent the intellectual property rights associated with works, known as
‘stuff’ in the model. This modelling tool forms the basis of ONIX, an e-
commerce system for the publishing industry which is covered later in this
chapter and described in greater detail in Chapter 9.
OAIS – Online exchange of data

The Open Archival Information System (OAIS) was developed to provide a
functional and information model of information preservation for access and
use by a designated community of consumers (Lavoie, 2004). The OAIS model
encompasses a range of information from that which is regularly updated, to
that which has periodic updates and from simple access systems to
sophisticated access systems and deals with the highly distributed nature of
many information systems. The simple model (Figure 3.9) is based on the idea
that information from a producer is input to the OAIS archive which is
managed and provides output for consumers such as the designated
community that it was intended for.
The information is packaged with a ‘wrapper’ of package information
about which description information is available. The package contains
Content Information and Preservation Description Information (PDI) (see
Figure 3.10). The PDI has a number of attributes (CCSDS, 2012):
• provenance – source of the content information, including its custody and

history
• context – relationship of the content to other information outside the
package
• reference – identifiers such as ISBNs (for books)
• fixity – which provides protection against undocumented alteration e.g.
check sum.
OAIS
Producer Consumer
Archive
Management
Figure 3.9 OAIS simple model
Preservation
Content Description
Information Information
Package Information
Figure 3.10 OAIS Information Package

DATA MODELLING 47
• access rights – terms of access for preservation, licensing, etc.
A slightly more complex representation of the OAIS model can be seen in

Figure 3.11, although this is still a high-level representation. The Submission
Information Package (SIP) is sent by the Producer to an OAIS archive. It will
contain some content and some Preservation Description Information (PDI).
One or more SIPs are transformed into an Archival Information Package (AIP)
which conforms to the internal architecture of the OAI archive. It will have a
complete set of PDI for the Content Information. When a consumer makes a
request for information the OAIS will produce a Dissemination Information
Package (DIP) in response.
Producer
SIP
OAIS
Query
AIP
Results
set
Order
DIP
Consumer
Figure 3.11 Relationship between Information Packages in OAIS
A fuller description of OAIS can be found in the ‘Reference Model for an

Open Archival Information System’ (CCSDS, 2012).
Conclusion
This chapter described how standards for metadata developed along different
paths to fit in with the requirements of different communities. A number of
data modelling systems or frameworks have been developed for describing
metadata.
The ABC Ontology is a general framework for developing domain-specific
descriptions and provides a way of describing different ontologies using a
common language. The Resource Description Framework (RDF) is a way of
modelling and describing metadata and can be expressed in a number of
languages, including HTML, XML, and Turtle (Terse RDF Triple Language).
Its syntax is based on triples, subject – predicate – object, so that, for example,
the book The Hound of the Baskervilles (subject) has creator (predicate) ‘Arthur
Conan Doyle’ (object). A third model, the Library Reference Model (LRM), is
more specific, providing a framework for describing products of intellectual
and artistic effort, such as books and sound recordings. Indecs, the fourth
modelling system described, focuses on the entities and transactions that
occur in a commercial publishing environment. OAIS, an information model
for digital archives, is used for online exchange of data.
CHAPTER 4
Metadata standards
Overview
This chapter looks at the structure of metadata standards. Understanding the way in
which standards are created and how they are constructed gives us an insight into their
use and potential applications to different situations. Metadata standards have arisen in
different collections and user communities and some of the main metadata standards
are described. This is not intended to be a comprehensive survey but rather an overview
of the range of metadata standards on offer with pointers to further information about
specialist standards. It starts with a description of Dublin Core, which is probably the
most widely used standard, partly because of its simplicity and partly because it was
designed for web resources. It illustrates some of the main features of metadata
standards and is itself used as the basis for specialist standards and application profiles.
The chapter goes on to consider some of the standards that are used in the library and
information field for bibliographic materials and social media. It also describes metadata
standards used for non-textual materials.
The nature of metadata standards

Metadata standards allow for exchange of data and by doing so ensure the
future usability of information resources. By having a documented standard,
it is possible for future users of the metadata to understand the conventions
in place and the intentions behind the metadata. Declared standards are an
important tool for controlling the quality of data about information resources.
Agreed conventions for generating content and for analysing resources in
order to describe them means that there is greater consistency. This allows
for interoperability, so that metadata can be exchanged between applications

and institutions. It also helps users by enabling them to confidently search
metadata collections, knowing where metadata such as dates, creator names,
and title information is held. Metadata can also be repurposed for use in new
environments and contexts. The growth of linked data has been an example
of extensive re-use of metadata collections. A model of the world and the
relationship between resources and other entities are implied by metadata
standards, which is why a prior understanding of data modelling (covered
in Chapter 3) is essential.
There is a process for creating standards based on international co-
operation and consensus. Although proprietary standards have persisted in
internal use by some database applications, most organisations that create
metadata use established, external metadata standards. In some cases they
may adapt existing standards to create their own application profiles. In the
UK, British Standard BS0 establishes the terminology of standard making and
provides a framework for development of national (and by extension
international) standards (BSI, 2011).
Standards arise in a number of ways. They may be created by an enterprise
and become a de facto standard for an industry. It is thought that the invention
of the standard screw gauge by Sir Joseph Whitworth in 1841 contributed to
the industrial revolution and was soon adopted in the naval and railway
industries (Galloway, 1958, 637–8). They may be imposed, for example by a
government agency, or they might be developed through a formal process of
consultations and consensus among interested bodies, as generally applies
for national and international standards. The International Standards
Organization defines a standard in the following terms:
A standard is a document that provides requirements, specifications, guidelines

or characteristics that can be used consistently to ensure that materials, products,
processes and services are fit for their purpose. (ISO, 2015)
The British Standards Institution provides an overview of standardisation,

including the terminology used and guidelines for standards development
(BSI, 2011). Although metadata standards are long-established in some
communities, in others standards have only started to emerge relatively
recently. Some domains, such as social media and digital humanities, are
relatively recent and the associated metadata standards may be less
established. Standards for non-textual material are also considered. For
example, there are established and emerging metadata standards for digital
images, movies and sound files.
METADATA STANDARDS 51
Specific standards are discussed in detail under the application area or

purpose that they are most closely aligned to. For instance, Chapter 7 on the
management of information resources includes a description of PREMIS, a
standard widely used for capturing the data associated with preservation of
digital materials. Another example of this approach can be seen in the
description of ONIX, which is widely used for exchange of bibliographic data
by the book trade, in Chapter 9.
About standards
Metadata standards provide a framework for analysing information resources
so that they can be effectively managed and easily retrieved. The standards
identify the characteristics of a resource that are recorded and sometimes
specify the way in which the metadata content is created (encoding schemes
are covered in Chapter 12). In effect the metadata standard provides
containers for particular types of information. For instance, in Dublin Core
the dc:creator data element, which is defined in the standard for Dublin Core
Metadata Element Set (ISO, 2009b), provides a place for author name, or
organisation responsible for the creation of that resource. The resource could
be a web page or it could be a book catalogued in a library, for instance.
Dublin Core – a general-purpose standard

The Dublin Core Metadata Initiative (DCMI) was set up in the 1990s following
a series of international workshops. It brought together members of the
library and computing communities in order to address some of the
information management issues arising from the emerging world wide web.
The initiative started in Dublin, Ohio, which is where OCLC (Online
Computer Library Center) an early supporter of this initiative, is based. Its
focus continues to be on web and internet usage and it is international in
scope. Annual meetings are held in different parts of the world and hosted
by participant member organisations (DCMI, 2015).
Although Dublin Core was set up to describe and improve the
discoverability of web and intranet resources, it has been widely adopted as
the basis for metadata in other domains, particularly in the library and
information field – to allow exchange of basic information and as a basis for
interoperability between systems (see Chapter 11 for a discussion of
interoperability of metadata). Because of its simplicity it is probably the most
widely used metadata standard and the core metadata element set has been
documented as an international standard (ISO, 2009b).
The DCMI Metadata Terms namespace includes the 15 core metadata

elements plus many others that are used to describe information resources.
Although originally designed for description of internet resources, DCMI
terms are now widely used for describing other information resources,
including printed materials, electronic documents and other digital resources
(DCMI, 2012). The data elements correspond to fields in a database or
catalogue, or properties in an RDF model. Although Dublin Core is a
permissive standard (i.e. it does not specify how the content of individual
fields is created), it recommends use of controlled vocabularies to generate
the content of the data elements. In order to allow backwards compatibility,
the standard is expressed in such a way that controlled vocabularies or
encoding schemes are not mandatory. The core data elements in alphabetic
order are:
• Contributor – This can be a person, organisation, service or software

agent ‘responsible for making contributions to the resource’. The data
type is ‘Agent’, an entity with the power to act.
• Coverage – If a resource is about a specific period in history (e.g. Roman
Empire, World War II, 1950s) or about a geographic or spatial location
(e.g. the dwarf planet Pluto, the State of São Paulo in Brazil, or the City of
Westminster in London, or the geographic co-ordinates 38.8977° N,
77.0366° W – also known as 1600 Pennsylvania Avenue, or The White
House), this can be recorded in the ‘coverage’ data element. This data
element can be qualified as dc:coverage.temporal or as
dc:coverage.spatial. The data element can also be used to indicate the
geographical extent of the applicability of the resource. For instance,
regulatory or legislative material may have a geographical extent.
• Creator – The agent responsible for the content of the resource is
normally the dc:creator. This may correspond to an author, or it may be
the job title of the person who wrote the content, or the organisation
responsible for the content of the resource being described.
• Date – The time or range of time during which a specific event in the
lifecycle of a resource takes place is held in this field. It can be qualified
to refer to a specific event e.g. dc:date.created. The event may be a future
event such as date due for review or disposal. The specification
recommends use of an encoding scheme such as W3CDTF to create the
date (see Chapter 12 for a discussion of encoding schemes).
• Description – This is text that summarises the content of the resource,
such as an abstract or table of contents. For web-based resources this is
sometimes taken from the opening paragraph of text on the web page.
• Format – This may refer to the file format (e.g. pdf file) for a digital
resource or the dimensions or medium of a physical resource (e.g. pbk 18

x 11 cm).
• Identifier – A (unique) reference to the resource is held here. Ideally this
should be part of a formal identifier system such as DOI, ISBN or URL
that uniquely and unambiguously identifies the resource (Chapter 5
describes resource identities and identifiers).
• Language – The language in which the resource is expressed is held in
this data element. The language can be described using standards such
as the ISO 639 series of two-, three- and four-letter codes (ISO, 2002). For
instance, pt or por for Portuguese. These can be combined with country
codes from ISO 3166 to produce more specific language designations
(ISO, 2013). So Brazilian Portuguese is represented by pt-BR and
Portuguese from Portugal by pt-PT.
• Publisher – The publisher is the agent or entity that makes the resource
available. This is usually an organisation name, but may be the name of a
service, or even an individual.
• Relation – This data element can be used to contain an identifier for a
resource that is in some way connected to the resource being described.
For instance, a web page may be part of a wider site. It is not to be
confused with ‘Source’.
• Rights – This refers to intellectual property rights such as copyright
ownership or access rights (for resources where access is limited to
specific groups). For instance, a Creative Commons CC-BY licence
(allowing use with attribution) would appear here.
• Source – If the resource being described is derived from another resource
or information entity, this data element can be used to identify where the
content comes from. This could be an identifier for a resource from
which the content of the resource being described is derived. For
instance, if the resource described is presenting data from another
publication, an identifier or bibliographic details of the original source
could be included here.
• Subject – The content of an information resource can be described using
keywords or classification codes. Dublin Core recommends that a
controlled vocabulary is used to obtain suitable subject terms. This could
be a general classification system such as UDC (Universal Decimal
Classification) or a subject-specific vocabulary such as MeSH (Medical
Subject Headings), or even a vocabulary designed for internal use within
an organisation.
• Title – The name by which the resource is usually referred is held in this
field. This would correspond to the title of a book or article, for instance,
or it might be the overall heading used for a web page e.g. ‘Google’ for
the Google search page, or ‘Library of Congress Home’ for the Library of
Congress home page.
• Type – The dc:type data element describes the ‘nature or genre of the
resource’. This might be: ‘web page’, ‘text’ etc. The file format or physical
medium would be described in the dc:format data element.
Some of the above examples show how a data element can be made more
specific by adding a qualifier. This is known as a ‘refinement’. So for instance
the dc:coverage data element can have a temporal or a spatial refinement,
appearing thus: dc:coverage.temporal or dc:coverage.spatial.
Different communities have emerged in the Dublin Core domain and they
have developed further data elements that extend the Dublin Core Metadata
Element Set (DCMI, 2012). Dublin Core is adaptable and supports the
development of application profiles. A Dublin Core Application Profile
(DCAP) incorporates elements of DC and data elements from other metadata
standards and namespaces or using specialised vocabularies to create a
standard for specific applications or requirements (Coyle and Baker, 2009;
Malta and Baptista, 2014). A DCAP is defined by:
• functional requirements
• domain model
• description set profiles and usage guidelines
• syntax guidelines and data formats.
The Singapore Framework (discussed in Chapter 12) shows the relationship

between these (Nilsson, Baker and Johnston, 2008). There are groups for the
DC-Library Application Profile and the DC-Education Application Profile.
Metadata standards in library and information work

We can see from the description in Chapter 1 of the origins of cataloguing in
ancient Middle Eastern civilisations that the library and information
profession has a long history of creating and managing metadata.
Resource Description and Access (RDA)

The introduction of the Resource Description and Access (RDA) cataloguing
standard has profoundly changed the way in which bibliographic data is
created, managed and used (Joint Steering Committee for Development of
RDA, 2014). Other long-established standards for exchange of library data
such as MARC have also played an important role in the development of
metadata standards for digital resources (Library of Congress, 2017c).

Chapter 3 describes the way in which the FRBR model (Functional
Requirements for Bibliographic Records, now a part of LRM) of bibliographic
data based on four levels of entity has been applied to information resources
(IFLA, 1998). The concept of Work, Expression, Manifestation and Item in
RDA provides potential benefits to the library community in more efficient
cataloguing and to users by delivering searches of related items. There are
data elements associated with each of these levels. The main purpose of RDA
is to support resource discovery of digital and non-digital resources. It also
incorporates the Functional Requirements for Authority Description, FRAD
(now a part of LRM), to describe person family, corporate body and place
(IFLA, 2013). In effect, RDA defines data elements for information resources
and the syntax for producing content (syntactic encoding scheme).The core
elements at manifestation and item level are (Joint Steering Committee for
Development of RDA, 2014):
• title
• statement of responsibility
• edition statement
• numbering of serials
• production statement
• publication statement
• distribution statement
• manufacture statement
• copyright date
• series statement
• identifier for the manifestation
• carrier type
• extent.
Up-to-date information can be found on the RSC website (RDA Steering

Committee, 2017).
MARC 21
The MARC standard (MAchine-Readable Cataloguing) emerged in the 1960s,
when libraries needed an efficient method of generating multiple catalogue
cards for each item. The advent of computerised processing of data allowed
for single entry of cataloguing details for multiple outputs for the author-title
catalogue and for the classified and subject catalogues. Individual catalogue
records were marked up with field designators to indicate the content of each
field. A number of national and specialist variations of MARC emerged and

ran in parallel until they were brought together in a single standard MARC
21 in the 1990s. It is based on US-MARC and is maintained by the Library of
Congress. As the website states, ‘The MARC 21 formats are standards for the
representation and communication of bibliographic and related information
in machine-readable form’ (Library of Congress, 2017c). MARC has been
influential in the development of other standards, originally designed to be
compatible with AACR2 cataloguing rules and compatible with RDA (Joint
Steering Committee for Development of RDA, 2015). The MODS standard
(described below) is a cut-down version of MARC 21. There are MARC 21
formats for:
• bibliographic data
• holdings data
• authority data
• classification data
• community information.
The focus in this chapter is on bibliographic data. The fields in the

bibliographic format metadata are labelled with numbers arranged in blocks,
as follows (Library of Congress, 2017c):
0XX = control information, numbers, codes

1XX = main entry
2XX = titles, edition, imprint
3XX = physical description, etc.
4XX = series statements
5XX = notes
6XX = subject access fields
7XX = name, etc., added entries or series; linking
8XX = series added entries; holdings and locations
9XX = reserved for local implementation.
The 1XX, 4XX, 6XX, 7XX and 8XX tags in the bibliographic format can be
modified by the following digits to give them a more specific meaning:
X00 = personal names

X10 = corporate names
X11 = meeting names
X30 = uniform titles
X40 = bibliographic titles
X50 = topical terms

X51 = geographic names.
For instance, this means that the tag 100 would contain personal names – such
as an author name, whereas 110 would indicate a corporate author.
100 1_ |a Amado, Jorge, |d 1912-2001.

110 2_ |a European Court of Human Rights.
Full guidance on use of MARC 21 for resources in bibliographic format can be

found on the Library of Congress Website: www.loc.gov/marc/bibliographic.
BIBFRAME
BIBFRAME, the Bibliographic Framework, is described in the following terms:
BIBFRAME provides a foundation for the future of bibliographic description,

both on the web, and in the broader networked world that is grounded in Linked
Data techniques. (Library of Congress, 2017a)
BIBFRAME is intended to address some of the shortcomings of MARC,

particularly with respect to RDA. Figure 4.1 shows the BIBFRAME 2.0 model,
Figure 4.1 BIBFRAME 2.0 model (source: Library

of Congress, 2017)
which describes the relationship between three levels of abstraction: Work,

Instance and Item. It also shows their relationships with the three core classes:
‘Agent’, ‘Subject’ and ‘Event’.
BIBFRAME uses RDF classes and properties to describe bibliographic
resources and the relationships between resources. The extensive BIBFRAME
vocabulary can be grouped in broad categories as follows:
• general properties
• category properties
• title information
• work identification information
• work description information
• subject term and classification information
• instance description statements
• instance identification information
• instance description information
• carrier description information
• item information
• type information
• cataloguing resource relationships – general
• cataloguing resource relationships – specific
• cataloguing resource relationships – detailed
• agent information
• administration information.
Metadata Object Description Schema (MODS)

MODS is a cut-down version of MARC 21 which has been developed for
resource discovery via the internet (Library of Congress, 2015b). MARC 21
records can be mapped onto MODS, although the reverse is more difficult,
because it involves parsing fields within MODS into separate fields in MARC
21.
The top-level elements in MODS are:
• titleInfo
• name
• typeOfResource
• genre
• originInfo
• language
• physicalDescription
• abstract
• targetAudience
• note
• subject
• classification
• relatedItem
• identifier
• location
• part
• extension
• recordInfo.
Some of these elements are container tags with no content, but serve to group
together sub-elements. For example <titleInfo> is a container tag with the
following sub-elements, which do contain data:
<title>
<subTitle>
<partNumber>
<partName>
<nonSort>
Individual elements or sub-elements may have attributes such as the

following:
Language-Related and Other Attributes: lang – xml:lang – script –

transliteration – altRepGroup – displayLabel
Date Attributes: encoding – point – keydate – qualifier
Linking Attributes: ID – xlink
For instance, the following MODS metadata refers to the English translation
of a text that was originally in another language (Portuguese):
<titleInfo xml:lang=‘en’ type=‘translated’>

<title>Gabriela, clove and cinnamon</title>
</titleInfo>
KBART
KBART (Knowledge Bases and Related Tools) is a link-resolving system that
enables libraries to link to appropriate copies of electronic publications such
as e-journal articles (NISO/UKSG KBART Working Group, 2010). The system
disambiguates references and links to licensed versions of documents where

appropriate. This means that a library user does not end up being directed to
a paywall for publications that his or her library already subscribes to. KBART
identifies the fields shown in Table 4.1.
Table 4.1 KBART fields
Label Field content
publication_title Publication title
print_identifier Print-format identifier (i.e. ISSN, ISBN, etc.)
online_identifier Online-format identifier (i.e. eISSN, eISBN, etc.)
date_first_issue Date of first issue available online
num_first_vol_online Number of first volume available online
num_first_issue_online Number of first issue available online
date_last_issue_online Date of last issue available online (or blank, if coverage is to
present)
num_last_vol_online Number of last volume available online (or blank, if coverage
is to present)
num_last_issue_online Number of last issue available online (or blank, if coverage is
to present)
title_url Title-level URL
first_author First author (for monographs)
title_id Title ID
embargo_info Embargo information
coverage_depth Coverage depth (e.g. abstracts or full text)
coverage_notes Coverage notes
publisher_name Publisher name (if not given in the file’s title)
General International Standard for Archival Description (ISAD(G))

ISAD(G) has been widely adopted as a basis for modelling, analysing and
describing archival materials (ICA, 2000). It is intended for use in conjunction
with local standards. It uses a multi-level model of archives based on Fonds,
Sub-fonds, Series, Sub-series and Items, and applies the following principles:
• description from the general to the specific

• information relevant to the level of description
• linking of descriptions
• non-repetition of information.
The elements of description (fields or data elements) that are applied to the
different levels of archives are grouped together as follows:
IDENTITY STATEMENT AREA

Reference code(s)
Title
Date(s)
Level of description
Extent and medium of the unit of description (quantity, bulk, or size)
CONTEXT AREA
Name of creator(s)
Administrative/Biographical history
Archival history
Immediate source of acquisition or transfer
CONTENT AND STRUCTURE AREA
Scope and content
Appraisal, destruction and scheduling information
Accruals
System of arrangement
CONDITIONS OF ACCESS AND USE AREA
Conditions governing access
Conditions governing reproduction
Language/scripts of material
Physical characteristics and technical requirements
Finding aids
ALLIED MATERIALS AREA
Existence and location of originals
Existence and location of copies
Related units of description
Publication note
NOTES AREA
Note
DESCRIPTION CONTROL AREA
Archivist’s note
Rules or conventions
Date(s) of descriptions
Although there are rules for the creation of the content of these data elements,
some of them are more narrative in nature and do not use controlled
vocabularies or specific syntactic encoding schemes. Elements such as
Reference code are based on international country codes followed by a
national reference number and a local reference code and there are clear
guidelines for generating the content of other fields as well.
The International Council on Archives established an Experts Group on

Archival Description (EGAD) in 2012 to reconcile four existing archival data
models (Gueguen et al., 2013).
Encoded Archival Description (EAD)

EAD is a detailed metadata schema maintained by the Library of Congress
(2016a). At the time of writing the third edition was being developed and
finalised by the Society of American Archivists. The standard provides a
detailed set of tags used for marking up different data elements associated
with archival materials. It is designed for use in a digital environment where
the metadata is handled by computer applications and is expressed in Relax
NG, XML and as a DTD. It is primarily designed to make archival materials
retrievable rather than for management or preservation. It is designed to be
compatible with ISAD(G).
Social media
There are several standards widely used for online and social media. Viewing
page information by right-clicking on a browser reveals structured metadata
embedded in the page. Some of the more common ones are briefly described
here.
FOAF
The FOAF (friend of a friend) ontology is:
a dictionary of named properties and classes [ . . .] FOAF integrates three kinds of

network:
• social networks of human collaboration, friendship and association;

• representational networks that describe a simplified view of a cartoon universe
in factual terms, and
• information networks that use Web-based linking to share independently
published descriptions of this inter-connected world.
(Brickley and Miller, 2014)
There are three types of entity in FOAF:
• Classes – groups to which other classes or agents belong

• Agents – entities that perform actions – they may be classes as well
• Properties – characteristics of an agent or a class.
FOAF statements in RDF define the relationships between entities. For

example, some personal information about a graduate of a university (or
‘school’) might include the following statements (in this example ‘me’ is
defined in a namespace such as http://mynamespace/info#me):
<foaf:Person rdf:ID=‘me’>
<foaf:name>David Haynes</foaf:name>
<foaf:title>Dr</foaf:title>
<foaf:givenname>David</foaf:givenname>
<foaf:family_name>Haynes</foaf:family_name>
<foaf:schoolHomepage rdf:resource=‘www.city.ac.uk’/>
</foaf:Person>
In this example the foaf:schoolHomepage would allow a link to be made to

other named individuals who have the same foaf:schoolHomepage.
Open Graph protocol

The Open Graph protocol is a Facebook metadata language for labelling web
resources so that they can be linked to other graph objects on the internet
(Facebook, 2014). The term ‘social graph’ (of which the Open Graph protocol
is an example) was coined to describe relationships in social networks
(Mislove et al., 2007). It is based on RDFa, which applies mathematical graph
theory to the analysis and mapping of relationships between individuals on
the internet. The nodes are equivalent to people and the edges (lines
connecting nodes) are equivalent to relationships. Facebook’s Open Graph
enables web pages to be turned into graph objects. A web page described in
Open Graph will have a minimum of the following properties:
og:title
og:type
og:image
og:url
And less-used data elements:
og:audio
og:description
og:determiner
og:locale
og:site_name
og:video
Twitter hashtags
Hashtags are widely used on social networks and microblogging sites by
users to enrich the metadata associated with a posting on social media. For
example, Twitter users can include hashtags in tweets to signify their content
or context. These are indexed by Twitter and are picked up as trending topics.
This allows a conference organiser to publicise a hashtag that all attendees at
a conference can use when tweeting about the conference. For example, the
ISKO UK biennial conference in 2017 had the hashtag #ISKOUK2017.
Although there is no formal control of hashtags on most social media sites,
they serve a useful purpose and continue to be a very popular way of
publicising events and marking specific topics for wider attention.
Non-textual materials
Standards such as MARC and MODS, and indeed Dublin Core, can be
applied to non-textual materials (Weber and Austin, 2011). However, there
are standards that have been developed specifically for non-textual materials,
such as VRA Core, MIX, and IIIF, which are described here.
VRA Core
The VRA Core describes works of art, cultural objects and their images, and
is maintained by the Library of Congress (Visual Resources Association, 2015).
VRA Core metadata can be embedded in METS (Metadata Encoding and
Transmission Standards) documents. The VRA Core has the following fields:
• work, collection, or image

• agent
• culturalContext
• date
• description
• inscription
• location
• material
• measurements
• relation
• rights
• source
• stateEdition
• stylePeriod
• subject
• technique
• textref
• title
• worktype
Metadata for images in XML (MIX)

MIX is one of a suite of metadata standards maintained by the Library of
Congress (2011). It is based on the NISO Z39.87:2006 standard for digital
images, which deals with technical image information rather than intellectual
property, provenance or preservation issues (NISO, 2006). For digital
photographs a lot of the technical image data is generated by the digital
camera and is captured in this metadata standard.
PBCore and EBUCore

Broadcasting services have developed several standards based on Dublin
Core, including PBCore in the USA and EBUCore in Europe (Corporation for
Public Broadcasting, 2011; EBU, 2015). PBCore is designed for audiovisual
material such as TV and radio programmes (Brighton, 2011). It is an extension
of Dublin Core and the data elements are grouped in four classes
(Corporation for Public Broadcasting, 2011):
• intellectual content
• intellectual property
• extensions
• instantiation.
EBUCore was developed by EBU, the European Broadcasting Union (EBU,

2015). Some work has gone into developing crosswalks between EBUCore
and PBCore so that data in one format can be converted into the other.
Moving Pictures Expert Group (MPEG)

Standards such as MPEG-7 are primarily for encoding digital content for
moving pictures, but include some metadata which is utilised in other
standards for audiovisual materials.
JPEG2000
JPEG2000 is a group of standards for image coding and compression. Part 2
of the standard deals with extensions to the coding data including metadata
associated with an image (ISO, 2004a). The metadata is divided into four
categories:
• image creation metadata – how the image was created, e.g. the camera and
lens settings
• content description metadata – the subject of the image, what it was about
• history metadata – what processing was done to the image to reach its
final form and or links to previous versions of the image
• intellectual property rights metadata – information about the rights owners,
etc.
EXIF
EXIF, developed by the Japan Electronic Industries Development Association,
overlaps in coverage with JPEG2000 but is independent of format (so that it
can also be used for TIFF image files, for instance, EXIF metadata can be
embedded in a JPEG or TIFF image. Figure 4.2 shows the general areas of
overlap between EXIF and two other commonly used metadata formats that
are used for images (Metadata Working Group, 2010, 21).
GPS
Orientation
EXIF
Copyright
Description
Creator
Date/Time
XMP IPTC IIM
Rating
Keywords
Location
Figure 4.2 Overlap between image metadata formats

International Image Interoperability Framework (IIIF)

The IIIF is an initiative by a group of research libraries and image repositories,
which was developed to improve access to image-based resources. IIIF
provides standards and APIs that allow digitised images and born-digital
image-based resources to be viewed by many different viewers. Examples
include the University Viewer developed by Digirati for the Wellcome
Library, Mirador Viewer, the British Library’s Georeferencer and e-codices,
the virtual manuscript library of Switzerland.
The IIIP Image API enables image servers to handle requests for images
and for information about images over the internet. It defines image request
parameters such as region (of the image), size, rotation, quality and format.
It also delivers image information, including technical properties and rights
and licensing properties.
The IIIF Presentation API enables servers and viewers to deliver images to
users, by focusing on the structural metadata. It is not intended for resource
discovery or content searching within a resource. A resource or digital object
may be described in terms of Manifest, Sequence, Canvas and Content (Figure
4.3). The relationships between these entities are shown in Figure 4.4.
Figure 4.3 IIIF object Figure 4.4 Relationships between IIIF objects (source: IIIF
(source: IIIF Consortium, Consortium, 2017)
2017
The resource properties are grouped under the following headings:
• descriptive properties
• rights and licensing properties
• technical properties
• linking properties
• paging properties.
The resource structure is described in Table 4.2.

Table 4.2 IIIF resource structure (source: IIIF Consortium, 2017)
Manifest The manifest resource represents a single object and any intellectual
work or works embodied within that object.
Sequence The sequence conveys the ordering of the views of the object.
Canvas The canvas represents an individual page or view and acts as a central
point for laying out the different content resources that make up the
display.
Image resources Association of images with their respective canvases is done via
annotations. Although normally annotations are used for associating
commentary with the thing the annotation’s text is about, the Open
Annotation model allows any resource to be associated with any other
resource, or parts thereof, and it is reused for both commentary and
painting resources on the canvas.
Annotation list For some objects, there may be more than just images available to
represent the page. Other resources could include the full text of the
object, musical notations, musical performances, diagram
transcriptions, commentary annotations, tags, video, data and more.
These additional resources are included in annotation lists, referenced
from the canvas they are associated with.
Range It may be important to describe additional structure within an object,
such as newspaper articles that span pages, the range of non-content-
bearing pages at the beginning of a work, or chapters within a book.
These are described using ranges in a similar manner to sequences.
Layer Layers represent groupings of annotation lists that should be collected
together, regardless of which canvas they target, such as all of the
annotations that make up a particular translation of the text of a book.
Collection Collections are used to list the manifests available for viewing, and to
describe the structures, hierarchies or curated collections that the
physical objects are part of. The collections may include both other
collections and manifests, in order to form a hierarchy of objects with
manifests at the leaf nodes of the tree.
Paging In some situations, annotation lists or the list of manifests in a
collection may be very long or expensive to create. The latter case is
especially likely to occur when responses are generated dynamically.
In these situations the server may break up the response using paging
properties.
The Advanced Association Features are:

• segments
• embedded content
• choice of alternative resources
• non-rectangular segments
• style
• rotation
• comment annotations
• hotspot linking.
IPTC Photo Metadata Standard

The IPTC Photo Metadata Standard contains three types of metadata (IPTC,
2014):
• Administrative – identification of the creator, creation date and location,

contact information for licensors of the image, and other technical
details.
• Descriptive – information about the visual content. This may include
headline, title, captions and keywords. This can be done using free text
or codes from a controlled vocabulary.
• Rights – copyright information and underlying rights in the visual
content, including model and property rights, and rights usage terms.
The standard is detailed with a core structure (data elements listed below)
and an extended structure with more detailed information on intellectual
property and some technical details.
City
Copyright notice
Country
Country code
Creator
Creator’s contact Info
• Address
• City
• Country
• E-mail address
• Phone number
• Postal codes
• State/Province
• Web URL
Creator’s job title

Credit line
Date created
Description
Description write
Headline
Instructions
Intellectual genre
Job ID
Keywords
Province/State
Rights usage terms
Scene code
Source
Subject code
Sublocation
Title
Complex objects
Other standards such as METS and OAI-PMH also allow for metadata
exchange between repositories and this is discussed in Chapter 6 on
information retrieval.
Metadata Encoding and Transmission Standard (METS)

Repositories and research libraries have to deal with complex objects in a
variety of formats. It can be difficult to capture all the necessary metadata
using a single standard and so the idea of metadata standards for complex
digital objects has emerged. Metadata standards such as METS act as a
container for metadata generated from other schemas (Library of Congress,
2015a). They may include specialist schemas for material in a specific format
or covering a specific subject-area. For instance, a METS record may contain
MODS metadata for internet discoverability, PREMIS metadata for
preservation information, and JPEG metadata for images that form part of
the resource. METS also allows for metadata standards to be used in parallel
so that, for instance, a METS record may contain both MODS and Dublin Core
metadata about the same resource.
One of the purposes of a METS document is to provide sufficient metadata
to allow a complex digital object (for example, an electronic document made
up of separate text sections, images and audio files) to be properly described
and rendered by a suitable software application or API. This allows for

exchange of digital library objects between repositories. A METS document,
expressed in XML, is divided into the following sections:
• METS Header – information about the METS document such as creator,

editor, etc.
• Descriptive metadata – this may contain metadata extracted from a MARC
record or Dublin Core record extracted from another application. This
section may contain parallel sets of metadata for the same resource.
• Administrative metadata – information about the digital object file,
including provenance, and intellectual property rights.
• File section – a list of the files that make up the digital object.
• Structural map – the structure of the digital object and links elements of
the structure and the content files.
• Structural links – records hyperlinks between nodes in the structural map.
• Behaviour – association between behaviours and content in a digital
object. This may include executable code to implement that behaviour.
OAI-PMH
The OAI-PMH (Open Archives Initiative – Protocol for Metadata Harvesting)
standard provides a framework for metadata discovery (Lagoze et al., 2002).
This enables service providers to ‘harvest’ metadata from other metadata
stores such as institutional repositories to create a searchable index and
repository. Some services also harvest content from other repositories to
facilitate faster retrieval. For instance, an institutional repository may collect
metadata in a variety of formats: authors inputting Dublin Core metadata,
MARC records from the library, LOM metadata from the institutional VLE
and DDIs from electronic publications. The institutional repository may also
make METS records available to external services Figure 4.5 on the next page
illustrates the process.
Dublin Core provides a ‘common currency’ for exchange of data about
internet resources. However, many consider it too crude for dealing with the
type of bibliographic material held in institutional repositories. MODS is
based on MARC and provides more detailed bibliographic data. It can be
generated from MARC records, such as those held on library management
systems, and so facilitate exchange of data between systems. Figure 4.6
illustrates the relationship between institutional repositories and resource
discovery systems. The service provider builds up a database of metadata
(and sometimes the resources themselves) harvested from institutional
repositories. It serves queries to the institutional repositories when an item is
identified during a user-initiated search. The provider then delivers the

retrieval item to the user.
Dublin Core
Institutional
Repository
Plus:
• Admin metadata
• Structural metadata
• Preservation metadata
Simple DC for harvesting METS

(OAI‐PMH) Objects
Figure 4.5 Metadata into an institutional repository
Figure 4.6 How OAI-PMH works
IEEE Learning Object Metadata (LOM)

Much of learning in higher education institutions (universities and colleges)
is delivered at least partially through virtual learning environments (VLEs).
Even in full-time courses with face-to-face lectures, tutorials and practical
sessions, blended learning is the norm. This means that students are exposed
to a wide range of digital materials in different formats.
IEEE LOM and its derivatives (such as ANZ-LOM) define data elements to
describe a learning object (IEEE Computer Society, 2002). These elements are
arranged into nine categories, as follows:
1 General – describes the learning object as a whole

2 Lifecycle – history and current state of the learning object
3 Meta-metadata – information about the metadata instance
4 Technical – technical requirements for the learning object
5 Educational – educational characteristics of the learning object
6 Rights – intellectual property rights and conditions of usage
7 Relation – relationships with other learning objects
8 Annotation – comments on educational use of the learning object
9 Classification – classifications applied to the learning object.
IEEE LOM metadata can be used in SCORM- and IMS-compliant digital

learning materials. SCORM (Sharable Content Object Reference Model) is a
set of standards and guidelines for learning objects that meet the following
functional requirements:
• accessibility
• interoperability
• durability
• re-usability.
The IEEE Learning Object Metadata standard (IEEE Computer Society, 2002)
and similar systems were intended to allow interchangeability of course
material. It was based on an assumption that it should be possible to construct
courses from pre-existing units and course material rather that writing from
scratch. In practice that objective has eluded educators. This may be because
course material is much more dynamic than many people acknowledged.
Most academics update their material at least annually. They also work hard
to make each course a coherent whole rather that an accumulation of
disjointed elements. Where things have changed is the growth of online
learning environments – particularly the freely available courses such as
Coursera, EdX and the Khan Academy. These are examples of MOOCs –
Massive Open Online Courses. Universities and consortia of universities have
set up these online courses, covering a wide range of subjects. A variety of
metadata on the courses is available for searching or display. This means it is
possible to search or navigate by subject, level and institution. Examination
of the landing pages of courses from the three largest providers reveals
extensive use of social media metadata such as Open Graph, Facebook and
Twitter. However, Miranda and Ritrovato (2014) identified Dublin Core, IMS
and IEEE LOM as widely used metadata standards for course material.
Conclusion
This chapter has provided an overview of some of the most commonly used
metadata standards in different domains of activity. It shows the relationships
between metadata standards and the domains in which they are used. The
choice of standards was also dictated by the use of widely accepted standards
as the basis for derived standards or application profiles based on national
need. Dublin Core was introduced as a general-purpose standard, even
though it was designed primarily to describe online resources and web pages,
specifically. We then considered standards that were applicable to LIS work,
such as KBART, RDA, MARC 21 and MODS and archives – ISAD(G) and
EAD. Metadata standards such as FOAF, the Open Graph Protocol and
Twitter hashtags were discussed in the context of social media standards. An
overview of standards for non-textual resources include VRA Core, MIX,
PBCore, JPEG and EXIF. Finally, complex objects that might include materials
in a variety of digital formats were covered by METS, OAI-PMH (for
exchange of metadata) and learning object metadata such as IEEE LOM.
Although some standards have persisted for a long time, Dublin Core since
2006, IEEE LOM since 2002 with minor amendments, other significant
changes are afoot. At the time of writing RDA had been implemented in
several national libraries, including the British Library, and the Library of
Congress. Other national libraries and many academic libraries are in the
process of implementing RDA (RDA Steering Committee, 2017).
Long-established mark-up systems such as MARC were under scrutiny
with the proposed BIBFRAME replacement being developed at the Library
of Congress.
PART II
Purposes of metadata
One of the organising principles of the first edition of this book was that
metadata could be categorised by purpose. The original five purposes
reflected the preoccupations of information professionals in the early 2000s.
Many of these purposes have stood up to scrutiny and Part II builds on that
model, but with six purposes. This part of the book starts with resource
identification and description (Chapter 5) as before. Chapter 6 looks at
information retrieval and the impact that metadata has on it. It necessarily
discusses retrieval theory moving beyond the measures of precision and recall
that were discussed in the first edition. This part then moves on to ‘Managing
information resources’ (Chapter 7) and looks at the role of metadata in
managing the information lifecycle. Chapter 8 considers intellectual property
rights, including provenance. Previously there was a chapter on e-commerce
and this has been developed into a description of the role of metadata
supporting e-commerce and e-government (Chapter 9). It is illustrated with
examples from the book trade (ONIX), e-learning environments and research
data (including ‘big data’). The final chapter in Part II is about information
governance (Chapter 10), dealing with ethical and regulatory issues. Risk is
used as a lens through which to view regulation and governance.
CHAPTER 5
Resource identification and description
(Purpose 1)
Overview
The chapter begins with a discussion of resource identifiers. It then considers how
resource description is used to distinguish between different information resources.
Some widely used identifiers such as ISBNs, DOIs, ISSNs, ISTCs and ISANs are described.
‘Description’ underpins other purposes such as retrieval and rights management. The
chapter then looks at other metadata used for describing information resources by
considering in turn: title; creator; bibliographic citation; date; format; and description.
How do you identify a resource?

Hider talks about metadata in terms of describing information resources, the
primary purpose of which is to facilitate information access and use. He goes
on to discuss data elements:
Each element describes an aspect or attribute of the information resource [. . .]

Clearly some attributes are more relevant in the provision of information access
than are others. (Hider, 2012, 15)
He makes the point that metadata used to describe a resource is a

representation of that resource. Taking this argument further, an identifier
can be seen as an extreme form of representation of a resource – reducing it
to a single sequence of digits or characters in a code. In effect an identifier
could act as a surrogate for the resource, although they are more usually
handled as labels for the resource.
78 PART II PURPOSES OF METADATA
Identifiers
A fundamental requirement of any description system is to have a way of
uniquely identifying an item, so that it is clear what is being described. This
is a particular concern in online records management, where there are
different levels of aggregation. Is the resource being described as a single
document, a series of documents on a particular topic, or a collection of items?
For a small collection, an identifier could simply be the title of a book or piece
of music. However, with even quite modest collections ambiguity becomes a
significant issue, as when two different books share the same title, or where
the same work may have different versions of the title (as with translated
works). Identifiers such as ISBNs can be used to distinguish between them,
although because ISBNs are assigned to manifestations rather than works,
they are not a reliable way of disambiguating two titles. Additional metadata
such as Author would be needed to distinguish between two works. For
instance, a search for the book title The Outsider in a public library catalogue
might retrieve the following items:
Albert Camus. The Outsider. (translation by Stuart Gilbert of L’Étranger

into English). London: Hamish Hamilton, 1946 – absurdist novel about a
murderer, set in French Algeria
Chris Culver. The Outsider. ISBN 9780751549126. London: Sphere, 2013 –
a detective novel set in the United States
Geordan Murphy. The Outsider. ISBN 9781844882793. Dublin: Penguin
Ireland, 2012 – autobiography of an Irish rugby player
Colin Wilson. The Outsider. ISBN 0753814323. London: Phoenix, 1956
(2001 printing) – a social psychology text on alienation
An ISBN alone may not be sufficient to identify an item. The first item in the
above listing predates the ISBN system, for instance. The last item was published
before the 13-digit ISBNs came into effect. It may also be necessary to distinguish
between several copies of a title in a lending library or between individual items
of stock in a publisher’s warehouse. An identification system can be used for
this as well. In both instances the identifier should be unique at some level (title,
edition, or item for instance) and unambiguous. Some works may have several
identifiers such as different identifiers for hardback and paperback editions of
a book. Translations present a particular problem because a translation could
be regarded as a separate work as well as an expression of the work in the
original language. It is important to understand what is being identified. The
FRBR model for bibliographic items allows for different levels of granularity of
information resource based on a multi-layer model comprising:
RESOURCE IDENTIFICATION AND DESCRIPTION (PURPOSE 1) 79
Works – e.g. The Dust Diaries by Owen Sheers

Expressions – e.g. the text of the work
Manifestations – e.g.
hardback edition, published by Faber & Faber in
2004. ISBN 0571210163
paperback edition, published by Faber & Faber in
2005. ISBN 0571210260
Items – such as particular copies of the
book in the British Library or in the Library
at SOAS, University of London, identified
by accession numbers
In the above example we see identifiers applied at each level from the author–
title catalogue entry for the work, to the manifestations of that work
represented by ISBNs. A digital object identifier (DOI) could play the same
role as identifier for a digital item. At the item level a library accession number
would identify an individual copy.
URLs (universal resource locators) are commonly used to identify web
pages; they are used throughout this book, for instance, to provide a reference
trail for those seeking further information or background about specific
topics. However, URLs describe the location of an electronic resource on the
internet. In most cases this happens to coincide with the actual resource and
so is effectively used as though it were a resource identifier. However,
websites change and the content at a particular address may disappear or be
replaced. This is one reason for giving the date accessed when citing a URL.
In other words, URLs are not necessarily persistent. The concept of a Uniform
Resource Identifier incorporates Universal Resource Names and Universal
Resource Locations (Berners-Lee, Fielding and Masinter, 2005). A URI may
be a URN that identifies a specific resource, but not how to access it, and/or
a URL such as a web address which points to a specific location on the
internet. In other words an ISBN and a URL are examples of URIs.
Name authorities – for many years libraries have developed name
authorities following AACR2, and now RDA (Joint Steering Committee for
Development of RDA, 2014). Archivists have also developed a system for a
name authority, ISAAR (CPF) for Corporate Bodies, Persons and Families
(International Council on Archives, 2004), which is used in several countries.
Name authorities ensure the consistency of catalogues and help to eliminate
ambiguity, one of the reasons for identification systems.
UUIDs and EANs

Two general systems have emerged for identification of items and objects, the
EAN (International Article Number) and UUID, the Universally Unique
Identifiers (ITU-T, 2014). Although not specifically designed for information
resources, they can be applied to digital and physical resources, such as e-
journals and printed books.
UUIDs operate on the basis of 16-octet numbers, the equivalent to a 128-bit
code. This yields such a large set of potential numbers to choose from that the
chance that any one number will be allocated to more than one object or item
is vanishingly small. The 16-octet number is represented by 32 hexadecimal
digits which are arranged in 5 groups separated by hyphens, making a total of
36 characters needed to represent a UUID. There are 2128 or 1632 possible
numbers to choose from. This is equivalent to approximately 3.4 x 1038 possible
numbers. Because there are so many UUID numbers available, they can be
applied to very large numbers of very small entities such as file locations on a
hard disc or individual devices on the internet. They can be applied to
ephemeral entities such as transaction IDs as well as to more persistent objects.
The standard includes a number of recommendations for algorithms for
generating UUIDs, to help ensure that they are unique. These include name-
based hashing, date- and time-based coding, and pseudo-random number
generation. An example of a numeric, pseudo-randomly generated UUID is:
a4739b4f-2077-4594-b8b3-2cfa70a41d5d
(Note: This code was automatically generated from the UUID Generator
website: www.uuidgenerator.net/. Each character (0-9 and a-f) represents
a hexadecimal (base-16) digit.)
EAN codes are used for physical objects and consist of a 13-digit code which
can be used to create a barcode (GS1, 2015). The EAN has retained its former
acronym, which stood for European Article Number, even though it is now
international in scope. These are the barcodes commonly seen on books and
magazines, and in the case of books correspond to the 13-digit ISBN, which
always starts with the 978 EAN code.
ISBN (International Standard Book Number)

The first ISBN standard was published by ISO in 1970 and since then ISBNs
have been widely adopted by the publishing, book and library sectors (ISO,
2009a). It provides a way of uniquely identifying monographs and other non-
serial publications (ISO, 2005). A different ISBN is allocated to each
manifestation of a title, so for instance the paperback version of a book will
have its own ISBN and the e-book will have a distinct and separate ISBN. For
example, each hardback edition of Harry Potter and the Chamber of Secrets
would have its own ISBN as would each paperback edition. However, ISBNs
can be mis-assigned or can be inadvertently re-used, so they are not always
reliable identifiers. ISBNs originally consisted of a 10-digit code. Since 2007
ISBNs have consisted of a 13-digit code with the following elements
(International ISBN Agency, 2014):
• EAN.UCC prefix – 3-digit number allocated by EAN International, which

is always 978 for ISBNs)
• Registration group element – which identifies the national, geographic,
language or other grouping within which the ISBN Agency allocating the
number operates. It is variable in length.
• Registrant element – indicates the publisher and varies in length according
to the projected output of the publisher
• Publication element – allocated by the publisher for the publication. The
length of this element will depend on the length of the registrant element
and the registration group that precedes it
• Check digit – based on a modulus 10 algorithm and providing a simple
way of checking the validity of a number, helping to identify
transcription errors, etc.
In order to manage the allocation of ISBNs, publishers are required to submit

ISBN metadata, including the following (International ISBN Agency, 2012):
ISBN
Product form
Title
Series
Contributor
Edition
Language(s) of text
Imprint
Publisher name and contact details
Country of publication
Publication date
ISBN of parent publication
Digital Object Identifier (DOI)

Digital Object Identifiers (DOIs) are used for identifying intellectual content
in a digital environment. DOIs are co-ordinated by the International DOI

Foundation via a network of national agencies. ‘The Digital Object Identifier
(DOI) was conceived as a generic framework for managing identification of
content over digital networks, recognising the trend towards digital
convergence and multimedia availability’ (International DOI Foundation,
2012). It is intended to form the basis for e-commerce. They are designed for
‘Interoperability with other data from other sources’, especially those related
to intellectual property items such as music recordings, written work or
museum artefacts (International DOI Foundation, 2012). The specification for
DOI can also be found in the ISO 26324 standard (ISO, 2012b). The
functionalities of DOIs is expressed in the following terms:
The DOI system offers a unique set of functionalities:
• Persistence, if material is moved, rearranged, or bookmarked;

• Interoperability with other data from other sources;
• Extensibility by adding new features and services through management of
groups of DOI names;
• Single management of data for multiple output formats (platform
independence);
• Class management of applications and services;
• Dynamic updating of metadata, applications and services.
(International DOI Foundation, 2012)
All DOIs start with the prefix 10, from the Handle System, followed by an
alphanumeric (letters and digits) of any length to identify the registrant
organisation (Sun, Lannom and Boesch, 2003). A forward slash separates the
prefix and the suffix, which is assigned to the entity or digital object itself.
The suffix may incorporate existing identifiers such as ISBNs. Once assigned,
a DOI is persistent – in other words it does not change, even if the ownership
changes.
The DOIs are based on three components: resolution, metadata and policy.
A DOI can be resolved into associated values such as URLs, other DOIs and
other metadata. A digital object with a DOI may have an associated URL, an
internet location (which is not necessarily persistent). The entity associated
with the DOI can be moved to another internet location or URL without the
need to change the DOI. The DOI can be resolved into multiple values, as we
see in the following example of four sets of associated data:
DOI:10.1004/123456
URL: www.pub.com
URL: www.pub2.com
DLS: loc/repository
International Standard Serial Number (ISSN)

ISSNs (International Standard Serial Numbers) are administered nationally
and co-ordinated by the ISSN International Centre in Paris. The standard, ISO
3297:2007, explains the syntax for ISSNs (ISO, 2007). The ISSN consists of an
eight-digit number, the first seven digits being the unique number allocated
to each registered serial title and the last being a check digit (based on a
Modulus 11 calculation). The ISSN is presented as two groups of four digits
separated by a hyphen, to prevent confusion with other international
standard numbers such as ISBNs. Approximately 2 million ISSNs had been
issued by 2017. Serials are registered with the relevant national ISSN agency,
which will require the following (ISSN International Centre, 2017):
title
frequency
publisher’s name
medium, etc.
ISAN – International Standard Audiovisual Number

The International Standard Audiovisual Number, ISAN, is a voluntary system
intended for use by the audiovisual industry to uniquely identify any work
(ISO 15706:2002 + A1:2008). It serves a number of purposes:
An International Standard Audiovisual Number (ISAN) identifies an audiovisual

work throughout its life and is intended for use wherever precise and unique
identification of an audiovisual work would be desirable. As an identifier, it may
be used for various purposes, such as to assist allocation of royalties among right
holders, to track the use of audiovisual works, for information retrieval and for
anti-piracy purposes, such as verifying title registrations. (ISO, 2008)
The ISAN consists of 16 hexadecimal digits, the first 12 of which are unique
to each audiovisual work and the remaining 4 being reserved for part
numbers. Machine-readable versions of the number have a check digit added.
The ISAN is applied to a work and all its manifestations in different media,
unlike ISSNs, which are unique to each form of a serial. A proposed
development is V-ISAN, which will incorporate information about the version
of the audiovisual work.
ISMN – International Standard Music Number

The ISMN, which corresponds to ISO 10957:2009, operates in a similar fashion
to the ISBN except that it is designed for handling notated music publications
(ISO, 2009a). It consists of a 13-digit code made up as follows:
EAN prefix – 979 for all ISMN items

Leading 0 followed by
Publisher ID
Item ID
Check digit
This conforms with the EAN system, which means that music publications
can have barcodes based on the ISMN.
ISTC – International Standard Text Code

According to ISO 21047:2009:
The ISTC provides a means of uniquely and persistently identifying textual

works in information systems and of facilitating the exchange of information
about those textual works between authors, agents, publishers, retailers, libraries,
rights administrators and other interested parties, on an international level.
(ISO, 2009c).
The code is made up of eight hexadecimal digits, as follows (International

ISTC Agency 2010, 9):
• registration element
• year element
• textual work element
• check digit.
The ISTC provides a way of identifying text which may be incorporated into
a serial publication or which may be manifest as a book. For instance, a work
with its own ISTC may correspond to several publications (manifestations of
the work), each with its own ISBN (Figure 5.1 opposite).
Archival Resource Key (ARK)

The Archival Resource Key is a system for persistent identifiers based on
URLs. An ARK reference takes the form shown in Figure 5.2 (Kunze, 2003).
Work by Author
with its own
ISTC #A
Manifestation of Manifestation of Manifestation of Manifestation of

Work as a Work as a Work as an Work as an
paperback hardback eBook html
ISBN #B ISBN #C ISBN #D ISBN #E
Figure 5.1 Example of relationship between ISTC and ISBN
http://ark.cdlib.org/ark:/13030/ft4w10060w
ARK Label Name (assigned

by the NAA)
Name Mapping Name Assigning
Authority Hostport Authority Number
(NMAH) (NAAN)
Figure 5.2 Structure of an Archival Research Key
RFIDs and identification

In contrast to ISBNs and some other identifiers discussed, RFIDs (radio-
frequency identifiers) can be used to identify individual items.
Radio-Frequency IDentity (RFID) tags were invented in 1948, based on the
idea that they can be attached to goods and materials and can be used for
identifying and tracking individual items. RFIDs are used in the retail
industry and have the potential to be applied to individual retail items,
components of those items, or at a macro level to pallets of those items, as
part of a logistics system, for instance. In libraries RFID tags are attached to
books, DVDs and other materials, and are used for security and circulation
control. With such a wide range of applications a number of standards have
emerged. The libraries version of the RFID standard is specified in ISO 28560-
2:2014 and is intended for use in all types of library (ISO, 2014b).
As RFIDs have developed, so their data-holding capacity has increased and
it is now possible to include significant amounts of data about the object to
which the RFID is attached. Where the object represents an information

resource, the data on the RFID is effectively metadata. RFIDs allow library
users to check out books for themselves, and to check them back in when they
have finished with them. The RFID code uses identifiers based on accession
numbers assigned to each individual item. Some libraries still rely on a
barcoding system for circulation control, and some use both: RFID for security
and barcodes for self-checking of items.
Describing resources
The six-point model of the purposes of metadata introduced in Chapter 1
started with resource description, which is the most fundamental of all
metadata purposes. It has its origins in the emergence of library catalogues
and at its most basic is a way of identifying works. Adequate description is
an essential prerequisite for information retrieval and resource discovery
(Chapter 6). It also underpins the other applications of metadata. Without a
way of identifying and describing a resource, it is impossible to use the
associated metadata for other purposes.
For example, in a web search (known generically as resource discovery or
information retrieval), some kind of description is needed for retrieved items
to evaluate the search results and to have an idea of whether the required
item has been retrieved. Another example would be in a library. A search of
a library catalogue that only yielded accession numbers would not be useful
for most searchers. Descriptive data such as the title, author or the format of
the item would normally be needed in order to evaluate the items and to make
a decision about its relevance and therefore whether to order, borrow, reserve
or consult the item.
A single data element may not be sufficient to distinguish between items.
A search for the author ‘Maya Angelou’, for instance, would probably bring
up several works. In order to select the appropriate item, a wider description
than just the name of the author would be needed to assess its relevance. The
book title I Know Why the Caged Bird Sings may then provide the additional
descriptive information that helps a reader to evaluate the retrieved item for
relevance and enables that person to distinguish between it and other books
by the same author such as Gather Together in my Name, or The Heart of a
Woman.
It may not always be clear how complete a description is needed in a given
situation. One extreme would be to use the entire item as the description. So
for instance, the entire text of a book could be used to describe the contents
of the book. In effect this is what happens with web pages or repositories of
electronic journals or e-books. The entire text is available for searching.
However, even this may not be complete, because it will not include metadata
elements that describe its context or what has happened to it during its life.
It also may not include external, independent descriptions of the item, which
may themselves be useful sources of data about the book, such as a critical
review, or a third-party abstract in a bibliographic database. The biggest
drawback of the complete text is its length – often making it impractical as a
source of information for rapid evaluation. This is why description metadata
is used as a surrogate for the full item.
Characteristics of metadata elements

There are many different systems for describing resources and they can be
grouped together by resource type, e.g. digital versus printed text, or text
versus images, or complex digital objects such as VLE resources. They vary
in complexity from the 15 data elements of the Dublin Core through to the
descriptions allowed for in RDA such as Title proper, Other title information,
Statement of responsibility and Summarisation of content.
Not all the data elements describe the intrinsic qualities of the information
resource. This section concentrates on those that do and looks at how they
handle resource description. The actual data elements used will depend on
the resources being described and might include bibliographic items
(electronic and printed), music, images, text, archives and virtual learning
materials. Intrinsic metadata is about qualities of the resource itself such as
its title, the author, and content description. On the other hand, applied
metadata may be contextual, describing how the item relates to other items,
or dynamic attributes such as ownership. The applied metadata may be about
the intended use of the resource. For example, the ‘Audience’ metadata
element defined by the Dublin Core Metadata Initiative shows that a resource
is targeted at particular groups, such as undergraduates, researchers or young
adults (DCMI Usage Board, 2012). Similarly, a record of transactions provides
another kind of contextual information which could have a bearing on the
provenance of a resource. Intrinsic attributes of identifiers are described in
the IFLA FRBR report in the following terms:
Attributes, as they are defined in the model, generally fall into two broad
categories. There are, on the one hand, attributes that are inherent in an entity,
and on the other, those that are externally imputed. The first category includes
not only physical characteristics (e.g. the physical medium and dimensions of an
object) but also features that might be characterized as labelling information (e.g.
statements appearing on the title page, cover, or container). The second category
includes assigned identifiers for an entity (e.g. a thematic catalogue number for a
musical composition), and contextual information (e.g. the political context in

which a work was conceived). Attributes inherent in an entity can usually be
determined by examining the entity itself; those that are imputed often require
reference to an external source. (IFLA, 1998)
Descriptive metadata
The following metadata elements (mostly derived from Dublin Core, with
the exception of Bibliographic Citation) are described in terms of their
relevance to describing resources. Dublin Core elements were chosen as the
basis for discussion in this chapter, because of its general nature, widespread
use and relative familiarity. It has been widely used as the basis for
application profiles relevant to specific communities of interest. The
descriptive metadata elements could include:
Identifier
Title
Creator
‘Bibliographic citation’
Date
Format
Description
Title
Although titles are extensively used to identify resources, they are not always
descriptive of the content. In web pages, apart from the URL the title is probably
the most widely used metadata element and in html it is delimited by the tags
<title> and </title>. This mark-up is frequently used by search engines and by
browsers to establish what is displayed at the top of the web page.
Book titles show considerable variation, depending on whether it is the full
title, a common representation of the title, or a particular translation of the
original title. This variation can cause confusion in the identification of an
information resource unless there is some way to distinguish between them.
Consistent cataloguing rules, for instance RDA and the ISBD, provide rules
on sources of information (Joint Steering Committee for Development of
RDA, 2014; IFLA, 2011). They can be used to establish which version of the
title takes priority, or even how to deal with different title origins. Cataloguers
have to deal with questions such as: ‘Does the title of a series of monographs
appear before or after the title of the individual monograph?’ They also have
to deal with subtitles and this may provide additional confusion in the
description of an item. For example, Charles Darwin’s famous work is often

referred to as Origin of Species. Yet when it was originally published in 1859,
its full title was: On the Origin of Species by Means of Natural Selection, or the
Preservation of Favoured Races in the Struggle for Life. This can cause problems
when searching for items in a catalogue and certainly causes confusion when
trying to determine which items are relevant.
In the discussion about RDA and LRM in Chapter 3, four different levels
of entity were identified, each with its own data elements, including title. A
work such as Origin of Species may be expressed as a written book which is
manifest in a number of different editions and reprints. In this case they all
share the same title and have to be distinguished by other means such as
edition, place of publication and publisher. These will be different
manifestations of the first edition, which is itself an expression of the original
work.
Work: The Origin of Species by Means of Natural Selection

Expression: Text of the First Edition
Manifestation 1: London: J. Murray, 1859
Manifestation 2: New York, NY: D. Appleton
and company, 1860
Thus, it is possible to see the benefits and limitations of title as a means of

describing information. In some contexts it may be sufficient to distinguish
between different items; in others it will be one of a number of attributes that
in combination provide a sufficiently detailed description of the item to assess
its relevance or to uniquely identify it.
Creator
Creator covers a wide range of possible relationships and may imply
intellectual property rights such as copyright. For printed publications, the
author is usually the ‘creator’ entity. However, it also applies to editors of
series of compiled works as well as illustrators and translators.
For web pages the situation can become quite complex. For instance, some
organisations do not attribute the content of web pages to named individuals,
but to departments or the organisation itself, often for the following reasons:
• to protect individuals against harassment – particularly if the content

may be viewed as controversial in some quarters
• to indicate corporate responsibility – especially for content written by an
individual or group of individuals in an official capacity
• to provide a more reliable point of contact for those who wish to act on
the content of the web page – the individual authors may move on, in
which case the department may be the more helpful point of contact.
RDA provides a good guide for expressing author names in publications. The
citation rules for many refereed journals also have their own conventions for
author names. The rules are not so clearly defined for web pages, and the
permissive metadata standards used in this arena such as Dublin Core do not
specify how a creator’s name should be recorded. The question arises should
it be: surname, followed by the initials, or full name, or the title followed by
the first name and then the family name? Even in the relatively well defined
area of bibliographic records there are variances in author name which can
cause problems when it comes to reliable identification of a publication.
Different amounts of data may be available for different publications by the
same author. In the previous example about Origin of Species, the author could
be expressed as:
C. Darwin
Charles Darwin
C. R. Darwin
Charles R. Darwin
Charles Robert Darwin
Charles Darwin 1809–1882
And then there are the inversions of the surname and given names: Darwin,
C. etc. The last example in the list above introduces the dates of Charles
Darwin’s lifespan, affording another way of discriminating between this
particular individual and other authors who may share the same name.
Where transliteration from a different script is concerned, there is an added
level of variation. For instance, is it ‘Tchaikovsky’ or ‘Chaikowski’? Is it ‘Mao
Tse-Tung’ or ‘Mao Zedong’?
A name may not be sufficient to distinguish between different authors. For
example, the author ‘Steve Jones’ comes in at least three distinct varieties,
which becomes clear when the cataloguing data reveals their dates of birth.
There is Steve Jones (b. 1944), the biologist who wrote The Language of the
Genes. Biology, history and the evolutionary future; there is Steve Jones (b.1953),
the sports writer and author of Endless Winter: the inside story of the rugby
revolution; and there is Steve Jones (b. 1961), the music critic and author of
Rock Formation: music, technology and mass communication. Authority lists which
include additional data such as the date of birth provide an added level of
specificity and makes identification of items (such as books) more reliable.
Where the creator is an organisation the issue of name change may arise.
For instance my own institution, ‘City, University of London’ was previously
‘The City University, London’ and prior to becoming a university was the
‘Northampton College of Advanced Technology’. These kinds of changes can
lead to problems with identification and accurate description.
Cataloguers deal with items ‘in hand’, so that the publication details reflect
the situation at the point of publication. This can be copy cataloguing, where
records are obtained from an external source and adapted, or original
cataloguing, creating records from scratch in-house (Chan and Salaba, 2016,
69). To some extent authority lists such as those maintained by the Library of
Congress (2017b) can help with making connections between different
manifestations or expressions of the same work. This is one of the issues that
RDA cataloguing is intended to address.
Bibliographic citation
The bibliographic citation includes elements already discussed, such as title
and creator. However it also includes other distinguishing details such as
publisher, place of publication and date of publication. Different types of
bibliographic records such as journal articles will also include details of the
journal and the volume and issue numbers of the journal. Conventions for
citations such as RDA or the multiplicity of conventions used by refereed
journals provide rules for the order and format of the citation details. The
intention of the citation details is to uniquely identify and help in the location
of the resource being described. Again, there must be a consideration of
consistency in citation conventions. Some applications such as RefWorks,
Endnote bibliographic reference management tool or applications such as
Zotero or Mendeley work with generic bibliographic records that can be
output in a variety of reference styles such as Harvard, or Modern Languages
Association, according to the requirements of the publisher. Even then these
conventions may be limited to the order in which items are cited and the
punctuation that separates the different data elements.
Date
Date information occurs in a number of contexts. It may be an intrinsic
property of an information resource – for example, date of creation, date of
publication, date of revision. It may also be an externally imposed data
element that has more to do with the management of the resource, for
example web page revision or expiry dates, or review dates for electronic
records. Date information can also refer to when something was done to the
resource – such as date of disposal, date of change of ownership. This type of

date information is not intrinsic to the resource or information container and
would not normally be considered to be resource description. However the
date of creation or modification is intrinsic and can also be used as part of the
identification and description of an item.
Format
Format information is particularly important for electronic information
resources and may provide the key to future access to the resource. It is
evident in digital images – many of which are created with a great deal of
proprietary metadata about: the format; the application and version used to
create or modify the image; the storage format; and the medium used to store
the data. This descriptive information becomes important when it comes to
reconstructing information by means of migration or by emulating the
original applications.
Format does not only apply to electronic resources. The format of a printed
work may also be relevant and may refer to whether a book is hardcover or
paperback, and the physical size of the document, and whether or not it
contains illustrations. This type of information is particularly helpful for
managing resources. Do the books fit on a standard shelf for instance, or do
they have to be kept with the outsize material?
Description
On the face of it the ‘Description’ data element (in Dublin Core) is most
directly relevant to the purpose of resource description discussed in this
chapter. However, descriptive information may not be an intrinsic property
of the information resource. An author’s abstract in a journal article or an
introduction from a monograph is intrinsic, but an externally produced
abstract or summary is not; it is applied to the resource. This becomes
particularly relevant in describing physical objects or images, as would be the
case in a museum.
There are different approaches to resource description. For example, an
external abstract may be enriched with controlled terms to enhance retrieval.
Alternatively, it may be purely free text – the most likely outcome of using
authors’ abstracts or publishers’ promotional material. The description will
depend on the purpose of the abstract and this will inform the approach that
should be adopted. Many secondary sources specialise in preparing abstracts
on indexed items. The same article may have quite different abstracts which
are geared to different audiences. The questions to ask are:
• Does it help users to assess the relevance of a retrieved item?

• Does it enhance retrieval – for instance by use of enriching controlled
terms?
• Is it intended to sell the item – as in publishers’ promotional material?
• Can it be used to evaluate a resource – is this an independent expert’s
commentary or review of the item?
The description data element can therefore be applied to this purpose even
though it is not necessarily intrinsic to the resource itself.
Conclusion
Description is an application of metadata that underpins other purposes,
including authenticity, finding and retrieving information and describing
what has to be managed. The actual names of the relevant data elements will
vary according to the schema used. Those that relate to description of
information resources and information-containing artefacts (such as museum
objects, digital media and printed documents) fall under the following broad
headings:
• Identifiers are a fundamental type of descriptor that allows for

discrimination between works or items being described. A number of
international standards for identifiers are widely used by a number of
metadata schemes. ISBNs, DOI and ISTC standards are all examples of
this.
• Titles are commonly used to identify resources such as web pages and
printed books. However, the way in which a title is expressed may not be
immediately obvious and this can cause problems in identifying relevant
items.
• Creator covers individuals and organisations and describes different
kinds of activity, from authoring a book through to composing music.
Names can be ambiguous and this has resulted in considerable effort
being devoted to developing authority lists.
• Bibliographic citations depend on consistent cataloguing rules and this is a
theme that is recognised as important for other data elements.
• Date of creation is intrinsic to many resources and is part of the
identification and description of an item.
• Format applies to both electronic and physical records and helps
managers to decide how best to handle an item.
• Description can be intrinsic (e.g. an author’s abstract or summary), or it

may be externally applied (e.g. a book review). Description helps in the
evaluation and selection of resources.
The data elements used to illustrate the descriptive purpose of metadata area
also fulfil other functions such as information retrieval, interoperability and
rights management. The level of description required will depend on the
context. For instance, a title may be sufficient for a library user to distinguish
between different books by an author. A fuller description (in combination
with other data elements) may be needed if several titles are being evaluated
to inform a purchase decision.
Identifiers are a particularly complex area, there being a variety of different
identification systems that can be applied. For instance, an electronic resource
may have a DOI, a URL and an ISBN. Other descriptive elements such as title
may be applied at the RDA work level or manifestation level. Throughout the
discussion on descriptive metadata elements, the theme of consistency has
recurred. The adoption of consistent cataloguing rules is one way of uniquely
identifying items and forms the basis for the development of authority lists.
In the library and archive fields there has been considerable progress in the
development of name authority lists that can be used to distinguish between
similar-sounding items and to consolidate variations around a party name
(such as an author) or information resource (such as an archive or book) for
consistent retrieval.
A common theme running throughout the description purpose is the need
for consistent encoding (which is covered in Chapter 12), to ensure a degree
of interoperability between items and to help discriminate between items that
are relevant and those that are not.
CHAPTER 6
Retrieving information (Purpose 2)
Overview
Metadata standards such as Dublin Core and MODS were designed to improve the
retrieval of web resources and discoverability of digital information resources. This
chapter considers the role of metadata in information retrieval. It begins with a review
of information retrieval concepts and measures of retrieval performance before
considering the impact of metadata on retrieval. Reference is made to models for
resource description and subject indexing. The final part of the chapter examines the
relationship between subject indexing and computational methods of retrieval.
The role of metadata in information retrieval

Van Rijsbergen (1979, 1–2) makes the distinction between information
retrieval and data retrieval. Information retrieval looks at the existence or non-
existence of a document or information resource that matches the search
criteria. This lends itself to document (information resource) descriptions, also
known as metadata. Data retrieval on the other hand is about obtaining
factual answers to a question, although the boundaries are being blurred with
the development of fact-based retrieval systems such as Google’s Knowledge
Graph. This can also be described in terms of metadata, although here the
emphasis is on data dictionaries that define the structure of the database
rather than describing document content. The focus of this chapter is on
information retrieval rather than data retrieval. This ties in with the overall
scope of this book on describing document content. However it does deal
with documents of all types from mainly text-based through to multimedia
materials. Metadata improves the discoverability of information resources by

describing the content in a variety of ways. Cochrane (1982) talks about
subject access as being systematic, topical or natural (free-text). The systematic
approach is via a classification or taxonomy which provides a formal
language for describing the content of the resource. The topical approach may
be subject headings which may be derived from a controlled language or may
be free text. The natural approach uses free text or natural language – i.e. text
retrieval from the content of the information resource itself. Considerable
effort has gone into text retrieval algorithms, primarily to rank the results in
a way that is meaningful and relevant to the searcher.
Automatic text analysis can be used for retrieval purposes. The frequency of
word occurrence in a set of documents provides a way of ranking search results.
Once stop words (such as articles and conjunctions) are stripped out, a plot can
be made of word frequency, i.e. words in rank order of frequency of occurrence
in the target document. The resolution power of words has been found to be
greatest among mid-ranking words. Once the most frequent words (which tend
not to be significant) and the least frequent words (which are too specialist) are
excluded, those in the mid-frequency range tend to be most useful for
distinguishing between documents during a search (Figure 6.1), (van
Rijsbergen, 1979, 16).
Upper frequency cut off

Resolution power of words
Lower frequency cut off

Number of words
Significant words
Words by rank order

Figure 6.1 Resolution power of keywords
RETRIEVING INFORMATION (PURPOSE 2) 97
General search engines have moved away from the classic Boolean search
model based on set theory, where exact matches to queries are required in
order to retrieve items. Algorithms based on probabilistic models allow search
results to be ranked by relevance (or closeness to fit). Although search engines
rank results so that the most ‘meaningful’ items appear at the top of the list,
they do not solve the problem of differing weights of search terms. Formal
descriptions extracted from the document or applied by cataloguers or
indexers still play an important role. This is particularly the case for the
semantic web, where the context of a descriptor can have a profound effect
on retrieval.
Another aspect to consider is the level at which retrieval takes place. This
can be at collection level or at the level of individual works, or individual
manifestations of that work, or individual items. Collection-level retrieval
provides ‘a filtering system that helps reduce users’ data overload’ (Zavalina,
2011, 105). The following types of data element had the greatest effect on
collection-level retrieval (in order): Description, Subjects, Title.
Most retrieval is document-based, but Salton, Allan and Buckley (1993) have
developed a passage retrieval system based on retrieval of excerpts of
documents rather than the whole document. This is particularly relevant where
a search yields many long documents which users then have to navigate
through to find the relevant material. This research has been carried forward
by exploring different computational techniques to improve the performance
of passage retrieval systems (MacFarlane, Robertson and McCann, 2004).
Passage retrieval can also be used to exploit the structure of XML documents
to narrow down the search results in XML element retrieval (Winter, 2008).
Information Theory
Shannon’s Information Theory, based on his work at Bell Labs, underpins
digital communications systems today (Shannon, 1948). It looks at the
probability that a particular unit of communication (such as a word or phrase)
will occur. The average quantity of information conveyed by a unit (expressed
as entropy) reaches its maximum when the probabilities of word occurrence
are all equal to one another. Otherwise, there is quite a lot of redundancy built
into most text-based systems. The less frequently a unit of communication
occurs, the more information it conveys. This can be used to compute the
incremental value of a two-word term over its separate components. In other
words, compound terms (or co-location of relevant words) can improve the
ranking of a retrieved document. This approach leads to a mathematical
analysis that is independent of linguistic analysis. The entropy, H, is equal to
minus the constant, k, times the sum of the probability of occurrence of a term,
i, times the log of the probability of occurrence of the term, i., expressed as:
Shannon’s Information Theory provides a measure of information in terms

of the probability of occurrence of a symbol (Shannon and Weaver, 1949). The
symbol might be a character, a keyword, or an indexing phrase. If the symbol
is an indexing term, we get a measure of how valuable or useful a term might
be for indexing purposes. Different techniques for term weighting have been
developed from this information theory. If the term appears in every
document in the collection it has no indexing value, because it does not allow
users to select sub-sets of the collection or to discriminate between individual
resources. If it appears in a few documents it could be useful, but it is difficult
to determine how useful it might be or whether there is an ideal frequency of
occurrence (Baeza-Yates and Ribeiro-Neto, 2011, 66–7).
Types of information retrieval

There has been a long history of information retrieval theory in the 20th
century, which is reflected in the development of many specialist search
engines and database systems. For very large document collections, ranking
of search results is critical to the utility of the search system and this has
shifted the emphasis of retrieval systems away from simple text retrieval
towards statistical approaches. These models can be applied to different types
of retrieval target (Baeza-Yates and Ribeiro-Neto, 2011, 59–61):
• Unstructured text
– Boolean (Fuzzy, Extended Boolean, Set-based)
– Vector (Generalised vector, Latent semantic indexing, Neural
networks)
– Probabilistic (BM25, Language models, Divergence from randomness,
Bayesian networks)
• Semi-structured text
– Proximal nodes
– XML-based
• Web-based
– Page-rank
– Hubs and authorities
• Multimedia
– Image retrieval
– Audio and music retrieval

– Video retrieval
The primary concern here will be with retrieval of unstructured text, such as
that typically found in books and journal articles. The description
‘unstructured’ does not mean that there is no structure to the text, but rather
that it does not conform to a standard structure (as defined in a document
type description or XML scheme, for instance). The text in a book will usually
be organised into a title page, contents section, chapters and sections and a
bibliography and index at the end. However these may not be defined in a
way that can be interpreted easily by a machine.
Semi-structured text such as that found in web resources is also of interest
as a great deal of retrieval by online search engines is based on this type of
resource. They exploit both embedded metadata as well as the text on the
page. HTML markers may indicate headings, allowing for a degree of
automation in the weighting of terms for retrieval. Titles and section headings
would tend to have greater significance than body text on the page.
The retrieval approaches can be broadly categorised as follows:
• Set Theory – such as Boolean logic and fuzzy searching

• Algebraic retrieval – such as vector spaces for ranking and latent
semantic indexing
• Probabilistic retrieval – such as Bayesian analysis
Boolean logic
Set theory has developed considerably since George Boole, a 19th-century
mathematician, invented Boolean algebra using logical operators to combine
sets. These basic operators are available on many search interfaces and are a
fundamental part of searching the internet and metadata collections such as
library catalogues. The commonly used operators are:
AND – both search terms or expressions linked by the ‘AND’ operator

must occur in the retrieved documents
OR – either search term or expression linked by the ‘OR’ operator must
occur in the retrieved documents
NOT – the retrieved set excludes all documents containing the specified
search term or expressions that follow the ‘NOT’ operator.
In Figure 6.2 on the next page a library catalogue contains details on books
about pets. In the first example an enquirer wants books about both cats and
dogs. The area of overlap between the two circles represents the set of books
on
Figure 6.2 Boolean operators
CAT AND DOG. Another reader might be less discriminating and may want
anything on either cats or dogs. This is represented by the total area of both
circles CAT OR DOG. In the third example, someone may be looking for books
that are exclusively about cats and which do not mention dogs at all: CAT NOT
DOG. This is represented by the left-hand circle, but excluding that part which
overlaps with the circle for ‘DOG’. Although this type of search facility is
available on many commonly used search engines, most users do not explicitly
use Boolean operators. They tend to be limited to advanced searchers. Google
and other search systems use the ‘AND’ operator implicitly to link two or more
search terms that are entered without operators between them. If it recognises
a phrase, it will be more specific than a single word. Other search engines,
particularly those found in intranets and on websites, use the OR operator by
default – expanding the search for each term that is entered into the query.
Fuzzy searching
Rather than a binary condition where a document is either a member of a set
or it is not, extended Boolean search models allow for weights to be attached
to terms and for a degree of membership of a set to be processed. Different
implementations of this approach have showed improved retrieval
performance over simple Boolean retrieval (Colvin and Kraft, 2016). So
document retrieval may be defined by the intersection and the union between
documents with respect to terms A and B:
Vector spaces for ranking

The move towards ranking of documents means that term frequency within
a document is also important. These measures have been incorporated into a
vector model that forms the basis of many subsequent vector retrieval
algorithms (Salton and Yang, 1973).
The frequency of occurrence of a term i in a document j is defined by the

function fi,j. The total number of occurrences of the term i in a document
collection of size N, is the sum of the occurrences in the individual documents
that make up the collection. When dealing with large numbers, it is more
convenient to use logarithmic values, because of the law of diminishing
returns. If Document A has 100 occurrences of a term and Document B has
1000, it does not mean that Document B is ten times as relevant. Following
this principle this equation can be modified as follows:
The Inverse Document Frequency (IDF) function, developed by Karen Sparck-

Jones (1972) measures the exhaustivity (number of index terms in a
document) and the specificity (number of documents which contain a term).
N is the total number of documents and ni is the number of documents that
contain the term ki.
Salton and Yang (1973) combine Term Frequency (TF) and IDF weights to
produce a vector value that is an indication of the usefulness of an indexing
term for retrieving and ranking search results.
Latent semantic indexing

An alternative approach is Latent Semantic Indexing, which is based on a
statistical analysis of the co-occurrence of terms in a retrieved set of
documents in order to refine the search results and to deliver a more reliable
ranking of search results. Commonly occurring words in a document are
identified as keywords of that document. Vector analysis is used to calculate
the similarity between documents and therefore cluster them for retrieval
(Baeza-Yates and Ribeiro-Neto 2011, 101–2).
Bayesian Inference and the Probabilistic Model

Bayes Theorem shows the relationship between two independent events A
and B. It states that the probability of A given that B is true can be calculated
from the product of the probability of A times the probability of B given that
A is true, divided by the probability of B:
The probabilistic model is an application of Bayesian Inference and provides

the basis for weighting individual query terms and documents. Bayesian
Inference is particularly useful because it provides a more accurate way of
estimating the probability of an event (such as whether a document is relevant
to a search query) where the probabilities are low. As the original paper says:
But what most of all recommends the solution in this Essay is, that it is compleat
in those cases where information is most wanted, and where Mr. De Moivre’s
solution of the inverse problem can give little of no direction; I mean, in all cases
where either p or q are of no considerable magnitude. [. . .] And tho’ in such
cases the Data are not sufficient to discover the exact probability of an event, yet
it is very agreeable to be able to find the limits between which it is reasonable to
think it must lie, and also to be able to determine the precise degree of assent
which is due to any conclusions or assertions relating to them.
(Bayes and Price, 1763)
Term frequency across a document collection provides a statistical method

for ranking documents. IDF and document length are also used in
probabilistic models. Where feedback is built into the system (either explicitly
where the searcher selects the most relevant items or implicitly by the system
monitoring use of retrieved items to assess relevance) the results can be
refined by a process of iteration.
Evaluating retrieval performance

Precision and recall
In text retrieval systems, retrieval effectiveness can be measured in terms of
precision and recall (van Rijsbergen, 1979, 148–50). These measures were
developed in the Cranfield experiments and are still widely used, notably in
the annual Text Retrieval Conference or TREC (National Institute of Standards
and Technology, 2017). In the context of the internet, these translate as two
performance issues:
• Lack of precision – internet searches often result in the retrieval of too

much material, most of it irrelevant
• Low recall – missing relevant material because the search is not
comprehensive enough or the resource is ‘hidden’.
Precision and recall

Precision is the proportion of relevant material actually retrieved to the total number
of documents retrieved.
(No. relevant documents retrieved)
Precision =
(Total no. of documents retrieved)
Recall is the proportion of relevant material actually retrieved in answer to a search

request to the total number of relevant items.
(No. of relevant documents retrieved)

Recall =
(Total no. of relevant documents)
These two measures can be expressed in terms of the following contingency table
where A is the total number of relevant documents in a set, and B is the total number
of documents retrieved:
Relevant Non-relevant
Retrieved A ∩B A ∩B B
Not retrieved A ∩B A ∩B B
A A N
A ∩B
Precision = B
Recall = A ∩B
A
Precision and recall are widely used for evaluating the effectiveness of
retrieval systems. As the precision increases, the recall often decreases. The
reverse is often true as well. As the number of items retrieved increases, the
precision (the proportion of relevant items in the retrieved set) decreases. In
practice precision and recall are difficult to measure, especially in a dynamic
and diverse environment such as the internet, because it is necessary to know
the total population of relevant items on the system. It can also be difficult to
assess the relevance of a retrieved item, especially if only one item is actually
needed to address the information need. In a web environment that uses
vector analysis and probabilistic searches to produce ranked results, it is
impossible to review the set of all documents that match a particular enquiry
in the Boolean sense. However the precision measure can be modified so that
a fixed number of ranked results is evaluated. For instance, if the first k items
in a retrieved set is evaluated it is possible to produce a measure such as the
precision at k. The first page of a search result may have ten items, and that
typically is as far as most searchers look. So an evaluation of those first ten
items makes sense to most users. This would be expressed as ‘precision at k
where k=10’ or ‘the precision of the first ten results’.
Another aspect of retrieval performance is recall. It is not always possible
to predict the effectiveness of a particular search query for retrieving relevant
items.
If different terminology is used by the searcher and the creator of the text,
there is likely to be a mismatch. For instance, a news report using the word
‘migrants’ may be about asylum-seekers or refugees as well as so-called
economic migrants. This might mean that a search on ‘refugees’ would miss
this news report. Controlled vocabularies and browsing systems may go some
way to addressing this issue, by associating related terms, or by providing a
navigation route to the preferred term. The role of thesauri and other
controlled vocabulary systems is discussed in Chapter 12.
Retrieval on the internet

Metadata can be used to put search terms into some kind of semantic context
– in effect telling the search engine or other application how to deal with a
particular metadata element. This is typically seen in a library catalogue,
where it is possible to distinguish between the author ‘Green’, (i.e. the name
of a person) and the keyword ‘green’, which describes a topic such as the
colour green. This kind of semantic distinction is becoming available on the
internet, using mark-ups of metadata (such as Dublin Core) embedded in web
pages. Metadata embedded in social media postings, for instance, are widely
used for selecting target audiences for e.g. online advertising. Algorithms
analyse metadata about user behaviour such as sites visited and interactions
with other users to group individuals into different categories for market
segmentation. Users can tag their own postings and indeed it is possible for
website owners to embed metadata in their pages. This is used on social
media sites where users’ descriptions are used as one type of information
retrieval. This is particularly notable in personal image collections such as
Flickr, SmugMug or Google Photos. User labelling extends to titles, galleries,
captions and tags (depending on which service is used). A collection of tags
on a service is sometimes called a folksonomy and raises many issues of
quality control and consistency. A more consistent approach would be to use
words selected from a controlled vocabulary such as a classification scheme
or thesaurus. This type of indexing can be facilitated by presenting the user
with a drop-down menu to select from, or a series of drop-downs to navigate

to the appropriate term.
Some web authors include the variant words or synonyms in the text of
their documents. This provides an alternative to explicit metadata elements
for subject content. However, this depends on the consistency with which
authors apply synonyms to the document and how comprehensively they do
it. A more consistent approach is to use controlled terms selected from a
thesaurus and to allow the search engine to automatically make the
connection between the controlled term and its synonyms, by means of
synonym rings (see Chapter 12).
Search engines and ranking

Search engines are constantly evolving and their approach to retrieval on the
internet has changed. The precise algorithms used by search engines tend not
to be published, because of the potential this gives web creators to manipulate
search outcomes. However, it is possible to see a pattern in the way in which
they have developed from simple pattern matching to sophisticated
algorithms that are based on probabilistic retrieval.
At the simplest level a search engine can simply look for all occurrences of
a word or a phrase in its database of internet resources. It may list the
retrieved items in the order in which it found the items. At a slightly more
sophisticated level it could list retrieved items in date order or in reverse
chronological order. This is manageable if a few items are retrieved, but not
particularly helpful for internet searches which can result in thousands or
even millions of hits. Given the general reluctance of users to deploy more
than one search term or to refine search strategies, some kind of relevance
ranking becomes essential.
Internet search engines started to develop ranking algorithms based on the
number of occurrences of a search term in the web page retrieved. Although
this can be helpful, it tends to favour longer documents where there is a
greater chance of the search term occurring multiple times. The next
refinement was to look at the frequency of occurrence of a search term. For
instance, 5 occurrences of the search term in a document 500 words long (a
frequency of 1 in 100) gives a higher frequency of occurrence than, say 10
occurrences of the same search term in a web resource which is 10,000 words
long (a frequency of 1 in 1000).
The big breakthrough was to realise that the position of a word, its context,
will have a bearing on its relevance. Users of library catalogues will be
familiar with the differences that occur when searching for a word in the title,
or in the text of a summary or abstract. A similar principle can be applied to
websites. If the word is in the title of a web page, search engines now tend to
attribute greater weight to it than if it only occurs in the main body of the
page. Matches to the title words push the resource up the ranking of hits.
Metadata can now play a role in putting a term into context. Unfortunately
this is a feature that was exploited unscrupulously by a minority of web authors
who embedded repetitions of keywords in the metadata. This manipulation
was carried to its logical conclusion by putting in the name of competitors in
the metadata fields of their home pages. This meant that searches for a
competitor’s name would retrieve the site indexed in this way - a good way of
alerting competitors’ customers to the existence of your products or services.
Because of the possibility of overt manipulation, most search engines reduced
the weight attached to metadata terms or ignored them altogether.
Search engines have continued to evolve and enhance the quality and utility
of search results by using semantic web features to make results more relevant.
Fact retrieval systems such as Google’s Knowledge Graph also depends on
semantic metadata. Ontologies, such as DBpedia, FOAF, schema.org and
Facebook Open Graph are all used to add meaning to search results. Metadata
is still important for retrieval, especially where users might want to restrict the
results by format, date or other criteria. Apart from social media sites, some
communities still add metadata to target pages to enhance retrieval. Domains
such as government or academic institutions may be more controlled in the
way in which they use subject terms to describe the content of their pages. For
instance, institutional repositories make subject (and other) metadata visible
by using Dublin core tags. This makes the resources discoverable by OAI-PMH
harvesting systems, which can then compile their own indexes of material (see
Chapter 4 for OAI-PMH). They regularly scan the target repositories for
updates which can then be incorporated into their own indexes.
Subject indexing and retrieval

The concepts of subject indexing can be applied to internet portals and
gateways and to intranets to enhance retrieval performance. Some of the data
elements used for resource description, such as ‘title’, ‘description’ and
‘creator’, are also used for information retrieval. Retrieval performance can
be improved by use of controlled terminology to describe the subject content
of the resource. Classification of material according to a taxonomy can also
provide a precise route in to relevant material.
Information retrieval in context

Intranets can be more tightly controlled than a group of websites, or the
internet. All too often intranets grow in an uncontrolled manner and do not
have a coherent structure. It is common for each department within an
organisation to have considerable autonomy about what goes on the intranet
and the result can be more like a scaled-down version of the internet than a
structured information resource. Content management systems can help
organisations to manage their intranets and websites more effectively.
However, even with use of the additional metadata elements to describe the
content of a particular page or site, there may be an issue of consistency.
Indexing resources is expensive in terms of human effort and lack of suitably
skilled staff can be a limiting factor. This absence can affect the quality and
consistency of indexing. Web managers need to be aware of these issues when
they are implementing a metadata strategy.
Using data elements to refine search results

Data elements provide a context for search results. One of the features of
metadata is that individual fields are labelled. This allows some search
engines and search interfaces to provide a ‘search by form’ or ‘query by
example’ approach. In the example in Figure 6.3 on the next page, entering a
term into the relevant fields makes it possible to retrieve relevant items. In
this example, from the British Library’s advanced search menu, the term
‘bean’ gives very different results if the search is in the Author field or in the
Main Title field. In each case the search is restricted to the appropriate
metadata element or cataloguing data field. This kind of approach is seen in
tightly managed systems where the contents of the information repository
can be controlled and structured with embedded data elements or are linked
to separate metadata records.
Metadata and computational models of retrieval

Human intervention is not always necessary for indexing. Many communities
of interest have explored methods of automatically indexing materials. In some
areas automated systems work in conjunction with human-applied indexing
held as metadata, in other areas it is seen as an alternative. Some of the
automated systems work with the content of the resource (different forms of
textual analysis), others with associated metadata (extracting terms from
description metadata elements) and others focus on analysis of the queries to
build up a user profile.
Figure 6.3 British Library search interface
Image retrieval
Multimedia files present challenges for retrieval because their content is not
composed of text which can be indexed and retrieved. Ponceleón and Slaney
(2011, 589) talk about the ‘semantic gap’ which they define as ‘the gap
between contents of a multimedia signal and its meaning.’ Retrieval by the
characteristics of the images or sounds can be achieved by a variety of
processing techniques and advances in face recognition and speech
recognition allow for subject retrieval in some cases. Content-Based Image
Retrieval (CBIR) has been focused on colour, texture and salient points of
images or multimedia files. This approach represents an alternative to
metadata-based retrieval.
The semantic content of images poses a particular challenge, because of the

need for human intervention, either at the point of creation of the multimedia
file (for instance by adding descriptive metadata to the file) or at the point of
ingestion of the multimedia file into a repository or database (for instance by
indexing or classifying images). There is still a requirement for human
intervention to analyse and describe the semantic content of multimedia
materials. There is a long tradition of organising and indexing audiovisual
material so that it can be retrieved by subject. Major broadcasters such as the
BBC have established archives of television broadcasts and many major
newspapers use their own or buy services from commercial image collections.
These collections are normally indexed manually by a variety of criteria
according to the likely requirements of their principal users. The indexing
may be very specific, such as ‘London Bridge’, or may be general – ‘bridges
carrying road traffic’. They may be indexed according to the predominant
colour – ‘red sunset’ – or by some generic abstract concept – ‘tranquillity’.
Chapter 4 discussed metadata standards, including MPEG-7, which captures
a variety of attributes about a multimedia object including subject. These
approaches can be developed and enhanced by deploying analysis of the
subject content in different ways.
MPEG-7 defines other attributes of images such as format of the image
file, resolution of the image, the application that originated the item and the
date of creation of the image, as well as the date of subsequent changes.
These are all criteria that can be used to select or to narrow down the
selection of images. In other contexts the same metadata elements are used
to manage the resources, described in Chapter 7. The formal metadata
standards will determine which information may be available. Many image
creation packages, such as digital cameras, attach their own metadata to the
image. This may be destroyed if the image file is transferred and saved in
another application. Other metadata associated with an image includes
format, resolution, originating application, date of creation and subsequent
changes.
The PhotoMetadata Project (Library of Congress and Stock Artists Alliance,
2009) identifies three broad categories of image metadata:
1 technical metadata, such as EXIF files, which are generated by the

capture device (e.g. camera)
2 descriptive metadata, such as the IPTC Core, which describes the
semantic content of an image
3 administrative metadata, such as the PLUS system, which deals with
licensing and intellectual property rights associated with an image.
The content-based image retrieval approach uses algorithms for automatic

processing of images to yield measures that can be compared with an example
image, or specified by a human operator (Hirata and Kato, 1992). It is an
alternative to concept-based searching that uses the textual content of
metadata associated with images. CBIR analyses images and matches the
resulting profiles against a query image to provide similar images in response
to a search. For instance, potential copyspace for use by advertisers can be
identified using CBIR. For other applications, some text-based indexing may
also be required – either by analysing the text surrounding the image, or by
explicitly applying indexing terms.
However, metadata still has a role to play:
The most important factor affecting what can be done with multimedia assets
(apart from their editorial value) is their intrinsic quality (e.g. the definition of an
image or the encoding format of a video) and the quality of the metadata
associated with them.
Metadata are textual descriptions that accompany a content element; they can
range in quantity and quality, from no description (e.g. Web cam content) to
multilingual data (e.g. closed captions and production metadata of motion
pictures). Metadata can be found:
1. embedded within content (e.g. closed captions);
2. in surrounding Web pages or links (HTML content, link anchors, etc.);
3. in domain-specific databases (e.g. IMDb1 for feature films);
4. in ontologies (like those listed in the DAML Ontology Library2).
(Ceri et al., 2013, 209)
Image and multimedia websites such as Flickr, iStock, Instagram and

YouTube use a variety of techniques for making their content available to
users. These exploit embedded metadata in image files, for instance, as well
as assigned indexing terms and user-generated tags. Figure 6.4 (opposite)
shows the metadata fields associated with an image in iStock, for instance
(iStockphoto LP 2017). This is reflected in the search parameters that are
available to users. The primary search window is for subject searching
(assigned categories plus tags assigned by the image owner). The selection
can then be refined by collection name, licence type, whether the image
contains people or not, the image shape (landscape, portrait, square) the
colour and the size of image. It is also possible to limit searches by format:
photos; illustrations; video; and audio.
Latterly, advances in machine learning mean that automated image
recognition is becoming much more sophisticated, with a range of techniques
for automatic image description and retrieval (Bernardi et al., 2016). Today there
Figure 6.4 Metadata fields in iStockphoto

(GettyImages, 2017)
are widely available apps on smartphones and other devices that allow users
to associate a face with a name. The system then automatically labels similarly
appearing faces in other photos, effectively recognising individuals in photo
albums. At the time of writing there were some experimental auto-captioning
systems such as CaptionBot which processes uploaded images and creates a
caption using artificial intelligence techniques (Microsoft, 2017). This approach
could provide enhanced retrieval from a collection of images that have been
processed in this way. Further work on deep indexing of images provides
more complex descriptions (Karpathy and Fei-Fei, 2017).
Conclusion
Although free-text searching and Boolean logic are powerful tools for
retrieval, more sophisticated statistical methods are widely used for internet
searching to provide a way of ranking search results. Shannon’s Information
Theory and Bayesian Inference have both played an important role in the
development of a new generation of search engines designed for handling
large data sets. The effectiveness of a retrieval set can be measured in terms
of relevance and recall, one of the fundamental developments in information

retrieval theory.
Metadata has a key role to play in high-quality information retrieval and is
particularly important in clearly defined domains. It also plays a key role in
providing users with options for searching on different attributes and for
putting the search queries into context. Despite these sophisticated options,
many users prefer to use a simple search window where a single term or
expression is entered. There is a great deal of potential for high-quality
retrieval on the internet, but a lot of this will depend on educating users about
the possibilities.
CHAPTER 7
Managing information resources (Purpose 3)
Overview
Management of information was the third of the purposes of metadata identified in the
six-point model of metadata use. This chapter describes the information lifecycle and a
simplified model of this provides the framework for describing the management of
information resources and the role of metadata. The chapter considers the role of
metadata in each of the main stages of the information lifecycle. This is illustrated with
examples from libraries, archives, records management and research data repositories.
Information lifecycles
One of the purposes of metadata is to manage the capture, storage,
distribution and use of information resources. This can be done in a variety
of contexts: libraries, records collections, archives, research data repositories
and multimedia collections. As well as formal collections, metadata also plays
an important role in the organisation of personal collections, such as
bibliographic references, social media and personal files.
The concept of an information lifecycle is widely used in the management
of digital resources and particularly for preservation. There are identifiable
stages in the life of a digital resource and this provides a basis for managing
those resources. Detlor (2010) puts forward a process-based view of
information management where the number of steps in the information
lifecycle depends on the perspective taken (organisational, library, or
personal). This is an idea that is also summarised in Floridi’s (2010) overview
of Information. Although metadata does not refer exclusively to digital
information (it is also used for books and other physical manifestations of
information) the majority of examples discussed here are electronic. These
include websites, electronic document and records management systems,
data repositories and social media. The lifecycle concept is well developed in
records management and this is reflected in ISO15489-1:2016, the international
standard for records management, which emphasises events in the lifecycle
of records:
Six broad classes of metadata may be used in the management of records. They
may be applied to all entities (see above), or fewer, depending on the complexity
of the implementation. The six classes are the following:
a) Identity – information to identify the entity;

b) Description – information to determine the nature of the entity;
c) Use – information that facilitates immediate and longer-term use of the entity;
d) Event plan – information used to manage the entity, such as disposition
information;
e) Event history – information recording past events on both the entity and its
metadata;
f) Relation – information describing the relationship between the entity and other
entities.
(ISO, 2016a)
Strictly speaking, record keeping could be described as a workflow, but it has

many features in common with the digital curation lifecycle. One of the
principles of records management is that: ‘records consist of content and
metadata, which describes the context, content and structure of the records,
as well as their management through time’ (ISO, 2016a). The standard
identifies the following stages in a record’s life:
• creating records
• capturing records
• records classification and indexing
• access control
• storing records
• use and re-use
• migrating and converting records
• disposition.
A record comes into being when a document is created or received and

attached to a registered file or given a file heading. During the capture process
MANAGING INFORMATION RESOURCES (PURPOSE 3) 115
metadata is created for the record and would typically include date of
creation, the owner and business classification. Metadata can be applied to
paper files or electronic records. Once created, a record may be retrieved and
used but not changed; this is known as fixity. It is classified and indexed to
ensure that it can be retrieved and that it is handled as part of the appropriate
class of records. During its life, access controls may be applied to a record so
that only authorised people are able to retrieve and use it. Preservation and
storage are important considerations, especially in a changing digital
environment. Disposal according to a retention schedule may be triggered by
an event, such as date of creation. The record may also be converted and
migrated to a new environment and this would be recorded in the record’s
metadata. ISO 15489-1:2016 specifies that metadata for records should include
the following:
a) a description of the content of the record

b) the structure of the record . . .
c) the business context. . .
d) relationships with other records and other metadata
e) identifiers and other information needed to retrieve and present the record . . .
f) the business actions and events involving the record throughout its existence . .
(ISO, 2016a)
Metadata standards such as ISAD(G) and EAD describe some of the data
elements that are used to manage records (ICA 2000; Library of Congress,
2016a). ISO23081-2:2009 provides a detailed metadata standard developed
for records management (ISO, 2009d).
The lifecycle model can be developed for digital curation of research data
such as that generated by research groups and typically held by universities.
Description and storage of research data sets has become an important way
of making research data available for further analysis. They may also be
consolidated into larger data sets. A simplified version of the Digital Curation
Centre’s (2010) lifecycle model is illustrated in Figure 7.1 overleaf.
Some of the attributes of this model can be simplified further to give us the
following defined stages in the life of an information resource or digital object
in a managed environment (see Figure 7.2, page 116):
• Create and ingest information – This is the original creation or compilation

of the information, its capture onto a system and attachment of metadata
such as index terms to that information source.
• Distribute and use – During this phase of the lifecycle the information is
accessed and utilised by a variety of consumers or designated
Concept-
ualise
Create or
receive Dispose
Appraise
Transform and select
Accesss Ingest
and use
Store Preserve
Figure 7.1 DCC simplified information lifecycle
Create or
ingest
Preserve
Transform
and store
Dispose
Distribute
Review
and use
Figure 7.2 Generic model of information lifecycle
communities of users. These may be specialist users, in which case the

management of their use may extend to issues of authentication and
rights management, or they may be general users, where the potential
uses may be difficult to predict.
• Review – An information source may be superseded or become

redundant. The review process is intended to ensure that the overall
information system continues to be up to date, relevant and accurate.
• Preserve and store – The integrity of the information sources needs to be
maintained with proper back-up procedures and migration policies as
the technology changes and develops.
• Dispose – Following the review process the information resource may be
destroyed, archived for future reference, or transferred to long-term
storage, where cheap storage and maintenance is offset against lower
accessibility.
• Transform – The record content may lead to creation of a new document,
which may be ingested into the system, and so the cycle begins again.
The remainder of this chapter will consider the role of metadata in each of
these stages in the lifecycle of an information resource, illustrated with
examples from librarianship, records management, digital curation and
content management.
Create or ingest
Different levels of aggregation will affect the ingestion of documents. For
instance, in records management systems, most of the metadata associated
with a record or file is generated at the point of creation or capture onto the
records management system. The metadata elements can be applied at
different levels of aggregation: an individual record (which may be an
electronic document or spreadsheet, for instance); at folder level
(corresponding to a paper file); or at class level (a file plan category). The
items at a lower level of aggregation inherit the attributes of their category,
so that for instance a document will inherit the metadata elements that apply
to the folder to which it belongs.
Institutional repositories often depend on data entry by the authors
themselves. This means that quality control issues may be a concern. Use of
drop-down lists can help to address some of these concerns, along with
intervention by cataloguers after the data has been entered.
Libraries have slightly different processes at the point of acquisition.
Notably, retrieval systems predominate and selection, ordering and purchase
of resources are significant factors. Co-operation between the book trade
(publishers, suppliers, retailers) and the library and information community
has opened up a number of possibilities. The acquisitions process can be
handled electronically – ranging from the small-scale ordering of individual
items via internet suppliers, or direct from publisher, through to purchase via
large-scale book suppliers, who may also select materials on behalf of the
library. Basic cataloguing data can be used to identify relevant items (e.g.
author, title information). If already known, identifiers such as ISBNs can also
be used for selecting titles. Publisher-supplied metadata can be made available
as part of the ONIX data, or MARC records can be located from central catalog-
uing agencies such as OCLC, the Library of Congress or the British Library.
The ONIX records allow for tracking of order information and verification of
delivery. ONIX is intended to support a fully-integrated e-commerce approach
to acquisition. This will include the delivery and payment details as well as
price and any discounts that may apply. Once the item has been acquired,
further cataloguing may be necessary to make it retrievable. Library
management systems usually import and export data in MARC 21 format,
although this may change with the introduction of BIBFRAME. Imported
records are enhanced with additional proprietary metadata from the library
management system and local data added by the cataloguers. The internal
metadata may include location information, loan records, details about the
management of binding of journals and covering of books and withdrawal and
disposal of items.
Preserve and store

The curation of information resources includes the storage and preservation of
items. In libraries, this is focused on the physical state of items. Digital items
such as electronic records and e-journals present additional challenges of
integrity, readability and obsolescence of technology and data storage formats.
The DCC Model allows for migration as a preservation action, leading to the
transformation of the resource and starting the lifecycle again (Digital Curation
Centre, 2010). This information is captured in PREMIS metadata, discussed
below.
One of the most useful reviews of metadata and preservation management
summarises four different strategies available for preservation of digital
resources (Day, 2004):
1 Preserving the technology – Preserving and maintaining obsolete hardware

and operating systems so that the original data media can be read on
them.
2 Emulation – Development of programs that mimic the obsolete systems –
so that the experience of using the data is as close to the original
experience but using up-to-date technical platforms.
3 Migration – Transfer of data to an up-to-date system – currently the
strategy adopted by many archiving bodies.
4 Encapsulation – Enclosing the original data with descriptive metadata

that allows it to be deciphered and viewed.
Surrogacy and conservation are two possible approaches to the preservation

of physical materials such as paper archives and printed books. Surrogacy is
important for physical resources such as incunabula (printed books from
before 1500) and illuminated manuscripts, which are fragile or which might
be damaged by further handling. Digitisation of the contents of a book or
archive, or the image of a museum object, has been widely used by libraries,
archives, museums and galleries worldwide. Commercial picture agencies
have also used this technology extensively. The focus is on images that can
be viewed and retrieved but not manipulated – in the case of documents, the
text is not encoded. With printed works, the text is often captured using
Optical Character Recognition (OCR).
BBC Domesday Book project

A digitised image or set of images will need to be managed in the same way
as encoded text and issues such as durability of the medium, the readability
of the file and the integrity of the image have to be managed if they are to be
accessible to future audiences. This has already become a problem for early
digital projects such as the BBC Domesday Book project, which was launched
in 1986 to coincide with the 900th anniversary of the production of the original
Domesday Book (BBC, 2017). The BBC project used what was the latest
technology of the day, a Phillips Laserdisc, to hold a combination of digitised
images and original digital content in the form of text, sound, stills and
moving images. The technology used then was specifically customised for
the project but is now obsolete.
One response to this challenge is to preserve the technology so that the
original media can be viewed in its original context. Preserving the original
technology is expensive and requires operational original equipment and
spare parts. It also requires the software to be maintained as closely as
possible to the original state. This approach does not guarantee that the
storage medium itself will maintain its integrity – so spare copies of the
storage medium (in the example described this would be the laserdiscs) and
the ability to write to them and read them (preservation of yet more
equipment) is required. The advantage of this ‘industrial archaeology’ is that
future users would be able to experience the digital images in a similar way
to the original audiences.
The alternative strategy is to migrate the file to a currently supported format.
In the case of the BBC Domesday project the content has been transferred to
DVD technology, making it available to a new generation of users. Metadata

on the migrated medium will help to ensure that the appropriate decisions
are made when it comes to the next migration. It can also be used to capture
data about the original format of the material prior to migration.
OAIS for preservation

The Open Archival Information System (OAIS), discussed in Chapter 3,
provides a functional model and information model for digital preservation
and is one way of handling preservation metadata (CCSDS, 2012). The
Archival Information Package defined in the OAIS information model
includes Preservation Description Information comprised of the following
types of information:
• reference
• provenance
• context
• fixity
• access rights.
It also provides a channel of communication between different communities

that are concerned with preservation issues, such as research data managers,
archivists, publishers and the library community.
PREMIS for preservation

The PREMIS metadata standard has been designed specifically to handle
preservation issues. It deals with core metadata that is likely to be used by
most repositories and can be extended to include data elements from
specialist schemes that deal, for instance, with rights and technical metadata
that ‘describes the physical rather than intellectual characteristics of digital
objects’ (PREMIS Editorial Committee 2015, 32). The PREMIS data model
establishes a relationship between Objects (information in digital form) and
the Rights (or permissions) associated with them and the Activities that have
an impact on them. Agents (people, organisations or software) may be
associated with Rights or Events but do not have a direct effect on Objects
(Figure 7.3 opposite).
1.1 objectIdentifier
1.2 objectCategory
1.3 preservationLevel
Rights Statement
Assertion of right or
permission
Object
Discrete unit of information Agent
subject to digital preservation. Environment Person,
Intellectual Entity, organisation or
Representation, File or software
Bitstream
Event
An action that
involves an Object or
an Agent known to
the system
Figure 7.3 PREMIS data model (based on PREMIS Editorial Committee, 2015)
1.4 significantProperties
1.5 objectCharacteristics
1.6 originalName
1.7 storage
1.8 signatureInformation
1.9 environmentFunction
1.10 enviornmentDesignation
1.11 environmentRegistry
1.12 environmentExtension
1.13 relationship
1.14 linkingEventIdentifier
1.15 linkingRightsStatementIdentifier
Wilson (2010) emphasises the importance of ensuring authenticity, reliability

and usability of research data over time and provides a critique of existing
standards such as PREMIS, which he considers insufficient on its own to
achieve these requirements and ensure the integrity of data objects. He makes
the point that research data managers can learn a great deal from the archives
community, which has pioneered the development of record-keeping
metadata standards for the management and preservation of digital objects.
Distribute and use

Software applications that manage information resources often use their own
proprietary metadata to enable the distribution and use of those resources.
Information resources include records, archives, library materials, research
data collections and web resources. The applications may also keep a record
of any transactions that take place – another type of metadata associated with
an information resource. This can be used to create an audit trail, which is a
useful element of information governance.
The loan of physical materials and access to digital resources generates
metadata about the resource and about the user and how the two interact.
The loan item metadata is normally derived from the cataloguing information
and will include data to uniquely identify each item in the collection,
equivalent to the item level in RDA (see Chapter 4).
Borrower details may include data about the number of other items
currently on loan. This is particularly true of academic libraries, where
different categories of user may have different borrowing privileges. Some
systems may keep a record of the loan history of patrons, although retention
of this data is controversial in light of anti-terrorism legislation such as the
Patriot Act 2001 in the USA, which allows the authorities to scrutinise loan
details of suspected terrorists. However, anyone can be a suspect.
Data protection legislation in Europe means that libraries are not permitted
to keep operational and transactional information relating to an individual
any longer than is required for its original purpose (European Parliament and
European Council, 2016). In this case the purpose would be managing loans
and return of items. Anonymised and aggregated loan data is used by
managers to assess the popularity of individual items and categories and to
inform decisions on future selection. The source of data about the borrower
usually needs to be verified, often by providing some kind of identification
and independent confirmation of address. Authentication is an important part
of managing metadata and is more fully discussed in Chapter 9.
For loan collections the location of a book is dependent on loan status and
is normally flagged up when the system is interrogated. This type of
information is made available to library users via the OPAC (online public
access catalogue) – sometimes with an indication of the likely wait time. The
screenshot in Figure 7.4 shows an example of a public library loan record with
the relevant metadata displayed, including due date, bibliographic details of
the item, the identifier (in this case a bar code number) and the status/action
date. Borrower details are displayed on a separate page.
Figure 7.4 Loan record from Westminster Public Libraries © 2017, Westminster City Council
Libraries and Sirsi Corporation
Review and dispose

Most collection managers are faced with review and disposal of less used or
out-of-date material. Metadata can be used to flag up items for review after a
set time. Enterprise search systems may allocate review dates at the point of
creation.
In records management systems retention periods are decided at the point
of creation of a record and may be associated with a business classification
system category or file plan category. Metadata can be associated with any
level of the file plan and lower levels inherit the attributes of the higher level.
For instance, if it is decided that all records relating to corporate policy-
making should be kept indefinitely, all the subsidiary records under that
heading would inherit this attribute. This makes management of records
much easier, because a decision is taken for a category of records rather than
on a record by record basis. There is still latitude to make exceptions where
necessary: so it may be that some of the corporate policy-making records are
ephemeral and need not be kept indefinitely, and they could have a much
shorter retention period.
Disposal decisions may apply to an individual record or a class of records.
Metadata associated with the folder to which the record belongs (or higher
aggregation such as a file category from the file plan) is also associated with
the record as it is captured or declared. This means that the disposal metadata
associated with the file category applies to the individual record as a default.
Depending on the standard, the disposal data element can be divided into
sub-elements to allow for effective management of the process. This will
include ‘Disposal action’, ‘Disposal time period’, ‘Disposal due date’ and
‘Disposal authorised by’. In this way the metadata can effectively trigger a
cascade of events during the record lifecycle.
Disposal is a necessary and often controversial aspect of library
management. Most library collections are living collections that are managed
with limited space available. Library managers have to decide how to
maintain their collections in a way that reflects the current needs of its users.
This means weeding out-of-date and damaged materials as well as acquiring
new titles. Metadata can be used to implement a retention/disposal policy.
Borrowing and usage patterns captured by the library management system
can reveal which items are not being used and which therefore may be
considered for disposal. Metadata associated with disposal will be of two
types – intention and action. Intention applies to documents that have a
known life or those that have been selected for disposal. Library management
systems can be configured to generate disposal lists (in much the same way
we have seen for records management), which can be reviewed before a final
decision is made. Once an action has been taken and recorded, the metadata
can provide an audit trail.
Transform
The transform step completes the information lifecycle, by leading to the
creation of new information resources. In creating a new document a new set
of metadata is required for each unique document. This may be embedded
within the document initially. New metadata is created when the document
is captured to the appropriate repository, whether it is a library management
system using MARC (or BIBFRAME) records, a records management system
using ISAD(G), or an enterprise search system based on Dublin Core with
added metadata elements. Research data is designed to be made available for
research and re-use. New data generated by manipulating or combining the
research data with other data sources leads to the creation of new resources
with their own metadata. Exposing metadata of existing resources through
exchange formats such as OAI-PMH, or by making it available in RDF format,
allows for the discovery, distribution and re-use. When combined with other
data sources a data set is transformed into something new.
Conclusion
Metadata is a tool for management of information resources, whether they
are electronic and available on the internet, via a closed system, or physical
and accessible via a library catalogue. Metadata enables lifecycle management
where resources are created, modified, used and disposed of. The metadata
is utilised by software applications to handle transactions. They also
document processes that have taken place during the lifecycle of an
information resource.
For example, records management systems depend on metadata to trigger
events in the lifecycle of a record. The metadata can also be used to anticipate
the fate of an individual record as soon as it is created, rather than after many
years when it is due for review and when the originators may have moved
on or retired. Another example of the use of metadata can be seen in content
management systems, which have their own metadata for describing and
manipulating web or intranet content.
Preservation is a complex area with a range of issues to be addressed,
including digital degradation and technology obsolescence. The use of
metadata becomes particularly important for digital materials because it
provides an avenue for describing the format and technology of a resource,
aiding its management and recovery. Metadata standards such as OAIS and
PREMIS are designed to facilitate preservation management.
Library management systems use acquisitions data to manage the
workflow from ordering and payment for a publication through to
cataloguing and making it available to users. Metadata associated with loans
keep track of individual items and assist with stocktaking exercises. Research
data collections also benefit from controlled metadata use to describe and
make available data collections for further research work, or for combination
with other resources as linked data. These examples demonstrate the wide
use of metadata for managing information resources. This is often based on
the management of a resource’s lifecycle.
CHAPTER 8
Managing intellectual property rights (Purpose 4)
Overview
This chapter considers the ways in which metadata has an impact on intellectual property
rights and information access rights. It goes on to describe the issues arising from
authenticity, ownership, and rights management. A discussion of different models of
intellectual property (IP) rights considers the Open Digital Rights Language as an example
of an information modelling language that deals with intellectual property rights. The
indecs system and PREMIS are referred to, because they both deal with rights, although
they are discussed in more detail elsewhere. There is also a brief discussion of the way
rights are handled by Dublin Core, MPEG-21 and the METS Rights extension to METS. The
chapter goes on to consider provenance, starting with a general definition and then
describing the PROV metadata standard. It then considers provenance in the context of
records management, e-documents, books and printed materials.
Rights management
Protection of intellectual property rights has a major economic impact on
many industries. One of the drivers for the development of metadata
standards in the publishing and book industry has been the need to manage
copyright effectively. They form a key part of the framework for publishing,
while protecting the rights of those involved in creating, performing or
distributing a creative work. In most countries an author has moral rights to
be identified as the creator of a work and consequently to enjoy the benefits
that come with these rights. The World Intellectual Property Organization
(WIPO) in Geneva regulates international treaties to help facilitate
international exchange and trade in intellectual property (WIPO, 2015). The

interests of different parties involved in intellectual property protection are
sometimes in conflict and use of the appropriate metadata helps to identify
those interests. Metadata provides a way of mapping the interactions between
the stakeholders and provides a mechanism for addressing rights and
ownership issues. The ease of copying and faithfulness of reproduction of
digital resources both pose an enormous challenge to publishers, record
producers and film makers. The growth of peer-to-peer servers for file-
sharing services continues to be a threat to copyright holders, because of the
way they bypass royalty payments. Wholesale breaches of copyright also take
place with electronic publications, software and other digital resources. The
challenges are two-fold. The first challenge is establishing ownership of the
rights, be they publishing rights, recognition of authorship or rights for
exploitation of the resource in new ways (such as a translation or publication
in a different format). The second challenge is ensuring that those rights are
applied and that conditions of use are not breached – or that if they are, that
the breaches can be detected.
Rights management metadata

Whalen (2016) identifies five groups of metadata elements that are to do with
intellectual property rights:
1 the name of the creator

2 the year the work was created
3 copyright status
4 publication status
5 date that rights research was conducted.
She goes on to tabulate some of the core elements for rights and this provides
a useful basis for identifying rights data in general catalogues or data
collections. Data elements dealing with rights are built into general metadata
standards such as Dublin Core or as extensions such as METSrights. For
instance, in Dublin Core the dc.rights data element provides a home for data
on copyright, licensing arrangements (such as those associated with Creative
Commons licences) and access rights (such as those invoked by freedom of
information legislation in different parts of the world). Dublin Core does not
specify how this data element should be expressed. Commonly the data
element is used to include a copyright statement to indicate ownership of
rights. Although the rights data element does not have any formal
refinements, individual authorities and organisations using Dublin Core have
MANAGING INTELLECTUAL PROPERTY RIGHTS (PURPOSE 4) 129
introduced their own refinements, such as Copyright, which would be

expressed as: dc:rights.Copyright. Another example of a refinement is
dc:rights.AccessRights to indicate intended audience and the conditions
under which they may access a resource.
PREMIS and intellectual property rights

Specialist metadata standards such as PREMIS (described in Chapter 7)
include optional rights data which can be linked to an information resource
or to an agent. For instance, what access rights or reproduction rights does a
repository have over a digital object in its collection? The rights statement and
links between the rights data and an object and/or agent help to define the
permissions associated with an information resource (PREMIS Editorial
Committee, 2015, 181):
4.1 Rights statement

4.1.1 rightsStatementIdentifier
4.1.2 rightsBasis
4.1.3 copyrightInformation
4.1.4 licenseInformation
4.1.5 statuteInformation
4.1.6 otherRightsInformation
4.1.7 rightsGranted
4.1.8 linkingObjectIdentifier
4.1.9 linkingAgentIdentifier
METSRights
METSRights is an external schema which is endorsed for use with METS
(Library of Congress, 2016b). The schema deals with intellectual property
rights associated with digital objects. It is an extension to METS and has the
data elements RightsDeclaration, RightsHolder, and Context. This allows
encoding of data about the nature of the rights associated with a digital object
as well as who owns the rights or has access for use of the digital object. The
Context provides a container for data about the circumstances in which the
rights apply and any constraints on those rights.
Open Digital Rights Language (ODRL)

Several models have been developed to help conceptualise intellectual
property rights and to provide a basis for the development of metadata
standards. These include indecs and the Open Digital Rights Language
(ODRL) and have led to the development of industry-specific metadata
standards such as ONIX (publishing industry), OAI-rights activity
(government, museums and libraries), and MPEG-21 (audiovisual materials).
Modelling systems and languages such as ODRL can be used to mark up or
express rights metadata.
The Open Digital Rights Language (ODRL) is described as ‘a standard
language and vocabulary for the expression of terms and conditions over
assets’ (Iannella, 2002). Terms and conditions include permission, constraints,
requirements, conditions, offers, and agreement with rights holders. ODRL
covers both physical manifestations and digital materials. It is an international
initiative to develop an open standard for digital rights management and is
designed to be compatible with a number of other models and standards for
rights management metadata, including indecs, EBX, DOI, ONIX, MPEG,
PRISM and Dublin Core. It provides cross-sectoral interoperability and is
extensible. In the ODRL language there are three core entities:
1 Assets, which are equivalent to resources, objects or intellectual property

and cover physical objects as well as digital content.
2 Rights cover the terms and conditions for use of the assets and will
include permissions, constraints, requirements for use and conditions.
3 Parties are equivalent to users or people and cover all types of roles from
end-user to rights holders and creators. The term ‘Parties’ applies to
organisations as well as individuals.
ODRL is a model that describes agreements between parties for rights over
assets and their use. The language can be used to model different types of
relationship and to allow for a range of interactions. The ODRL Foundation
Model is illustrated in Figure 8.1 (Iannella, 2002).
Permissions cover four areas of activity: usage, re-usage, transfer and
asset management. Within each area a number of specific activities are
described:
• Usage includes permission to display or print a resource (such as text),

play a video or music or to execute a program.
• Re-usage covers permission to modify content, excerpt material, annotate
items and aggregate the resource with other resources.
• Transfer covers permissions to sell, lend, give or lease items. These are
commercial in focus and would cover the permission of libraries to lend
books, for instance.
• Asset management covers in-house management of the resource so that it
Figure 8.1 ODRL Foundation Model © 2002 World Wide Web Consortium (MIT, ERCIM,
Keio, Beihang)
can be installed, backed up, moved, deleted and restored. These are non-
trading activities, but are necessary for effective maintenance of the
resource within a client organisation.
ONIX
The book trade provides a good example of the complexities that arise when
it comes to managing intellectual property rights. Rights-related metadata
includes information on: authorship, publishers and territorial rights. The
ONIX metadata framework was developed with this partly in mind
(EDItEUR, 2014). In order to develop ONIX a framework was needed to
analyse the different types of relationship that occur and are necessary for
commercial transactions to take place. The indecs model is just such a
framework, developed with support from the European Commission with a
focus specifically on rights management (Rust and Bide, 2000).
By establishing rights ONIX allows for automated rights management and
for the use of rights while protecting rights owners and allowing freedom of
legitimate, fair use. There are different views of metadata, including the
intellectual property view (Figure 8.2):
make
persons intellectual property
used by
own in
rights
Figure 8.2 Legal view of entities in ONIX
• persons make intellectual property

• intellectual property is used by persons
• persons own rights in intellectual property.
An entity, such as a person, may have different attributes depending on the

view used. In this chapter we have looked at only one view of the relationship
between persons, intellectual property and rights. Each of these entities has
an identity and a function. In the indecs model, intellectual property is
recognised as a legal concept defined by national legislation and international
agreements and treaties. Indeed, different types of intellectual property are
defined by treaties such as the WIPO Copyright Treaty (WIPO, 2017) and the
WIPO Performances and Phonograms Treaty (WIPO, 2017). Indecs actually
uses metadata to describe ownership of rights etc. and it provides a data
modelling language from which application profiles such as ONIX can be
developed. A more detailed description of indecs can be found in Chapter 3.
MPEG-21
The MPEG-21 series of standards provides an interoperable framework for
multimedia. The aim is to work across a range of communities and to facilitate
integration of different models. MPEG-21 encompasses content creation,
production, delivery and consumption (ISO, 2004b). It defines a framework
for Intellectual Property Management and Protection (IPMP). The purpose is
to enable legitimate users to identify and interpret intellectual property rights,
whilst enabling rights holders to protect their rights. The Digital Items
Declaration Language (MPEG-21 DIDL) is an interoperable schema for
declaring digital items. The language can be used to represent the Digital Item
Declaration model and is one element of MPEG-21.
Rights management as an enabler

Although there are systems focused on rights management itself, it is mostly
seen as an enabler that allows trade to take place. By establishing the
ownership rights for intellectual property, and in particular digital objects,
the rights of creators and producers can be protected and this gives them an
incentive to produce and release new products onto the market. For instance,
the Creative Commons (2016) makes use of embedded metadata to facilitate
the use of copyright material without the intervention of intermediaries (see
Figure 8.3). This means that an author or creator of material can make it
available on the internet with standard licensing conditions. The metadata is
generated from a website app and includes details of the licence type and can
be embedded in a web page:
Figure 8.3 Creative Commons Licence, licensed under CC BY 4.0
<a rel=‘license’ href=‘http://creativecommons.org/

licenses/by/4.0/’><img alt=‘Creative Commons Licence’
style=‘border-width:0’
src=‘https://i.creativecommons.org/l/by/4.0/88x31.png’
/></a><br />This work is licensed under a <a rel=‘license’
href=‘http://creativecommons.org/licenses/by/4.0/’>Creative
Commons Attribution 4.0 International License</a>.
Additional metadata can be added to the embedded metadata, such as the

author, title of the work, attributions and other permissions, as well as the
format of the work. The XML/RDFa code generated by the Creative Commons
website can be embedded in the web resource or other digital material. The
title and source metadata are marked up as Dublin Core terms.
Provenance
Provenance – the place of origin or earliest known history of something.
(Pearsall, 1999)
When it comes to establishing the authenticity of an item, its history becomes

important, its provenance: the circumstances of its creation, who owned it,
and the conditions under which its ownership was transferred.
Records management and good governance depend on being able to
demonstrate the authenticity of a record and to provide documentation about
its history and the way it has been managed. This may include details of
transactions that have taken place: who viewed a particular document and
when; what changes were made to the document during its history; and the
measures adopted to ensure that unauthorised changes have not taken place.
This provides the basis for legislation on the legal admissibility of electronic
documents and whether they can be used as evidence in legal proceedings.
Provenance is something that has dominated trade in the art world. The idea
that a painting is what it purports to be (for instance knowing who it was
painted by, when, that it is not a forgery or copy) will affect its perceived value.
This idea also applies to printed books and other physical artefacts, where there
may be a value associated with an original manuscript or a first edition. This
idea has been adopted in the commercial world and applies to documentation
associated with business transactions. In the context of digital materials,
providing provenance information can help to demonstrate that a record has
not been tampered with and that the evidence that it presents is therefore
reliable. Moreau (2010) surveys the literature of provenance on the web and
concludes that the benefits of publishing provenance data outweighs the costs.
Metadata can provide a record of the provenance of a document and
evidence that it has been kept to set standards and following defined proced-
ures. This is vital for documents that have been scanned and digitised and
where the original has been destroyed, as well as born-digital documents. Many
banks and building societies routinely scan and digitise financial documents.
The metadata associated with the digitised image helps to ensure that the
resulting image is legally admissible in court. Traditionally, the authenticity of
documents with legal weight such as contracts and wills was established by
the signature and an identifying mark such as a seal or watermark. They also
had metadata associated with them, such as details of how the document had
been kept and information about the procedures in place to prevent

tampering or indeed changes of any kind to the original.
In a document produced using an application such as Microsoft Word, for
instance, the system updates details of the editing process and can keep track
of version numbers, providing an audit trail. Identity numbers associated
with electronic documents, digital images and other electronic resources such
as DOIs (Digital Object Identifiers) are an essential part of any system that
purports to preserve authenticity.
PROV is a standard for provenance metadata, which is hospitable to
provenance metadata from other schemas. It is based on a model of Agent,
Entity and Activity, shown in Figure 8.4 (W3C, 2013).
Figure 8.4 PROV metadata model for provenance © 2013 World Wide Web Consortium,
(MIT, ERCIM, Keio, Beihang)
There are different approaches to recording the provenance of an item,

depending on the type of item being described:
• records and archives

• digital images
• electronic documents
• books and printed material
• metadata.
Records and archives

In records management, knowing who has had responsibility for a document
(especially an electronic document) is a part of the control of its integrity.

Being able to develop an audit trail is a key aspect of good governance. The
concept of respect des fonds, the idea of grouping archives according to the way
they were created or respect for the origin of the archive, is a part of archival
theory which can be used to explore the issue of provenance. Being able to
trace records, or scientific data (from a data archive) back to its original source
is facilitated by metadata.
Digital objects
Legal admissibility of digital objects such as born-digital documents and
digitised images depends on the accompanying documentation and
certification attesting to its authenticity. Metadata provides a way of recording
details of the circumstances of creation of a document (date of creation, author,
editor, etc.) and actions that have taken place since – an audit trail of who has
accessed the document and any changes that have taken place – what the
amendments were and who made them, when. As discussed in Chapter 2,
many programs automatically attach their own metadata to electronic docu-
ments when they are created and this can provide an audit trail for the
document as it is drafted and altered.
There is no way to verify the authenticity of a document without
information about its history and what has happened to it since its creation.
For this, metadata is necessary. The authenticity of information can be
determined by means of physical certificates to indicate that the document
has been checked or that a specific procedure has been followed, or via the
metadata embedded in the resource or held separately in a database.
Books and printed material

Bibliographic records can incorporate details of ownership and past history
of printed documents, whether they be archives, or formally published
material. Some standards allow for user-defined fields for details of
ownership and circumstances of change of ownership – for example, sale of
an item. This type of information applies at item level (in FRBR terms) and
tends to be used only for special items such as first editions of famous works,
older material and items that are important because of who owned them –
for instance, books with dedications or inscriptions by famous people inside
them. Older materials such as incunabula (pre-1500 printed materials) and
illuminated manuscripts need details of their provenance to help verify their
authenticity. If it is not possible to account for its complete history, fraud
becomes a real possibility. This is an issue that museums and art galleries face
with individual paintings and objects in their collections. Good metadata used
in conjunction with other tools and scholarship to establish the age and origin
of an item help to build the case for the authenticity of an item.
Metadata
Provenance is a key aspect of the preservation data associated with an
information resource. Metadata and metadata schema can be treated as
information objects or digital objects for the purposes of describing
provenance and managing their preservation. This allows the application of
preservation models such as PROV and PREMIS to metadata (Li and
Sugimoto, 2014). This approach would help to record data such as who
created the metadata record, what content rules were used to create it, and
when was it created or amended.
Conclusion
In looking at rights management we see some similarities between different
models of intellectual property rights. It quickly becomes clear that there are
three main concepts that need to be represented in any model of rights
management:
1 The item, content or resource – the intellectual property.

2 Agent, party, person, or organisation – this entity can play a number of
roles from intellectual property owner to consumer and any of the
intermediaries.
3 Rights – including the terms and conditions of use as well as details of
ownership and other relationships between the item and the agent.
The models are capable of a great deal of complexity as we have seen above,
but the use of comparable building blocks allows for a degree of inter-
operability between different schemas arising from the models.
Rights management metadata was developed in response to the need to
protect the intellectual property rights associated with digital resources and
a need to allow for the different types of transaction that take place in creating
and distributing electronic resources. In order to do this, models for
intellectual property rights (IPR) management such as ODRL and indecs were
developed. Specialist metadata schemas such as PREMIS, MPEG-21 and
METS Rights are also used for capturing and handling rights data for digital
objects. Another aspect of ownership is provenance, which can affect the
acceptance of the authenticity of an item and therefore its value. It is also
important in controlled environments where an audit trail of transactions may

be required. The PROV metadata schema provides an avenue for handling
provenance of digital objects including metadata and metadata schema.
CHAPTER 9
Supporting e-commerce and e-government
(Purpose 5)
Overview
This chapter considers the ways in which metadata is used for e-commerce and e-
government. It describes use of metadata for marketing and online behavioural
advertising. E-commerce is illustrated with an example from the book trade, ONIX, and
with a description of music industry metadata and digital images. It finally looks at e-
government, focusing on the documentary aspects of transactions and the role that
metadata plays in facilitating these transactions.
Electronic transactions
E-commerce and e-government are two sides of the same coin. They are about
human interaction with organisations via the internet that result in
transactions of one kind or another. In the case of e-commerce that interaction
is with commercial organisations. In e-government the interaction is with
public bodies. E-government has a slightly wider definition, in that public
education and information can also be considered a part of e-government
even if it is not a part of a specific transaction.
The main difference between the two is that transactions with individuals
may involve metadata behind the scenes, but do not require overt handling
of metadata by the consumer. For instance if an individual is recording a life
event via the internet (such as registering a death), the ‘Tell Us Once’ service
in the UK effectively allows surviving relatives to complete one death
registration form and the data on that form is used to inform the local
authority, the tax authorities, benefits agencies, vehicle licensing, passport

office and national savings accounts.
Many e-commerce and e-government systems depend on tagging
(metadata) to make their sites discoverable by target groups and the use of
metadata in this way is explored in Chapter 6 on retrieving information. E-
commerce and e-government systems also deal with personal data and data
about website interactions using technologies such as cookies to track usage
of a site. This raises privacy issues which are discussed in Chapter 14. This
chapter includes a discussion about metadata for cookies in the context of
online behavioural advertising, a major focus for e-commerce activity.
E-commerce
Chaffey (2015, 13) defines e-commerce in the following terms: ‘e-commerce
should be considered as all electronically mediated transactions between an
organisation and a third party it deals with.’ Laudon and Traver (2014, 51)
talk about ‘the use of the Internet, the World Wide Web (Web) and mobile
apps to transact business’. E-commerce now plays a role in most businesses
in their transactions with customers and with other businesses. As well as
direct retail activities, businesses procure services and purchase products
from suppliers using e-commerce applications.
Metadata plays a key role in the revenue-generating activities of social
media giants such as Facebook, Google and Yahoo!. For instance, van Dijck
(2013, 63–4) says: ‘As Facebook owns an unprecedented reservoir of
customised (meta)data, advertising and public relations are becoming a
mixture of science and statistics, and therefore a lucrative business model.’
This points to the enormous potential being realised by control and
management of metadata associated with use of internet resources. Talking
about another major platform, he continues (2013, 93): ‘Needless to say, both
user-added tags and automatic tags added considerably to Flickr’s
commercial potential, especially in the area of app development and
recommendation systems.’
Search engine optimisation

The header of a web page will often contain meta-tags (i.e. metadata) that
describe different attributes of that page. This information is used by browsers
to help them present the pages in an appropriate way to users. Meta-tags are
also used by search engines for display in search result listings and may also
be used for ranking the search results. The metadata found in the headers of
web pages may include the following:
SUPPORTING E-COMMERCE AND E-GOVERNMENT (PURPOSE 5) 141
• title
• description
• keywords
• robot behaviour (index, noindex, follow, nofollow).
Image tags may also be used for retrieval, and certainly for display. An
‘alternative description’ is commonly used so that an image description is
given when a cursor is rolled over the image – a feature originally intended
to help users with visual impairments. However, this Alt text provides a
textual description of an image which assists retrieval as well. Metadata about
language or country may also be useful for global sites with different
interfaces for different groups. For instance, invoking www.google.com will
deliver the user to an appropriate country site, based on the IP address of the
user. This may be important where there are different products available,
different pricing, or different regulations that apply to transactions in each
country. So, for instance, following the European Court of Justice ruling on
Google (Spain), all the EU versions of Google provide a disclaimer at the
bottom of the results page if it detects that you are doing a search on an
individual, allowing it to comply with a European Court of Justice ruling in
2014: ‘Some results may have been removed under data protection law in
Europe.’
Some search engines, such as Google.com and Bing.com, are set up to look
for marked-up meta-tags in the items and to generate snippets for display in
response to searches. So-called ‘rich snippets’ may be used to improve search
results or enhance the display of search listings. These may include
breadcrumb trails, ratings reviews, pricing and meta-descriptions. Rich
snippets are a form of semantic mark-up metadata in web pages that allow
search engines to interpret their content more accurately and deliver relevant
results pages to users. Google uses Schema.org (described in Chapter 12) as
a standard for marking up semantic data.
Online behavioural advertising

One of the major developments of the past ten years has been the growth of
online behavioural advertising (OBA). This is probably the dominant feature
of income generation on the internet. Although the sums associated with
individual transactions may be small (cents or fractions of a cent) the
enormous volume of traffic means that major internet presences such as
Google (via Alphabet, its holding company), Facebook and Yahoo! are now
among the largest global corporations. The techniques used for tracking usage
of websites and clicks-through to target sites have evolved and are still
evolving, partly in response to regulation and partly to reflect changing user

behaviour. In the early 2010s Facebook and others were using beacons, small
pieces of code embedded in websites that could be placed within the user’s
browser. This beacon is then detected and followed by the advertising agent.
Cookies are another technology prevalent at the time of writing. These pieces
of code contain some basic metadata (which varies according to the type of
cookie). They may document things such as what sites have been clicked on,
a timestamp for the click, as well as the last time the user was on the web and
when the cookie was last invoked.
Figure 9.1 shows a simplified information flow that occurs during a
browsing session with visits to two sites that are part of the same ad network.
In the figure a user visits website A (step 1). The website places a cookie on
the user’s browser (step 2). Later, the user visits website B which is part of
the same ad network (step 3). The cookie is detected by the ad network server
(step 4). The ad network server compares the user to an ad profile and the
appropriate ad is delivered to the user via website B (step 5). Activity
associated with the cookie is also recorded as part of that user’s profile.
Figure 9.1 Cookie activity during a browsing session
Different types of cookie metadata are available, depending on the browser.

For instance, Mozilla Firefox (version 44) shows the following metadata
elements for session and persistent cookies:
• Name – filename for the cookie

• Content – a unique ID or name – value pair
• Host – domain where the cookie applies
• Path – specific area of the domain where the cookie applies

• Send for – allows for encrypted connections for sensitive data
• Expires – this is a timestamp indicating how long the cookie persists for.
If this value is blank, it is a session cookie which is deleted at the end of
the session (i.e. when the browser is closed).
The Maximum Age is also sometimes included with some browsers to

distinguish between session and persistent cookies. If cookies are
unencrypted or are encrypted using standard encryption techniques, they
may be susceptible to hijacking and fraud. In some cases it may be possible
to gather up the transactional metadata associated with a cookie (credit card
numbers, passwords to secure sites, personal contact details).
Indecs and ONIX

Bide (2011) talks about ‘the development and implementation of
communication standards in support of e-commerce: essentially, identifiers
and metadata.’ The indecs metadata framework described in Chapter 3
provides a way of modelling e-commerce transactions. Chapter 8 went on to
describe indecs in the context of intellectual property rights. E-commerce
involving intellectual property requires effective rights management to be in
place.
The indecs model states that ‘People make stuff’ and this is the starting
point for rights management and e-commerce. The indecs model states that
‘People use stuff’, which means that there is a market for intellectual property,
including information resources. The third statement ‘People do deals about
stuff’ is an acknowledgement that trade takes place. Although the indecs
model is intended to deal with commercial transactions, it also deals with
transactions such as lending from a public library. In the indecs model a
library lends books (stuff) to library members (people) free at the point of use.
Money transactions do of course occur at some stage; the library has to buy
books and in some countries royalties may be payable to authors when their
books are lent out.
In order to manage the metadata associated with e-commerce transactions,
the indecs framework puts forward the following principles, which could be
applied to other domains as well, and are therefore worth rehearsing here
(Rust and Bide, 2000):
• Unique identification – ‘Every entity should be uniquely identified within

an identified namespace.’ This ties in with the axioms that metadata is
modular, and that stuff is complex. Without a unique identifier for each
entity it becomes difficult and expensive to administer some aspects of

the e-commerce chain. The principle goes on to describe the attributes of
an identifier of uniqueness, stability, security and public availability of
basic descriptive metadata for the entity identified.
• Functional granularity – ‘It should be possible to identify an entity
whenever it needs to be distinguished.’ This means that an item should
be capable of being identified at whatever level of granularity is
appropriate. For a library this may be the entire collection, or for a music
recording it may be a compilation of many pieces from different artists.
However for royalty payments, it may be necessary to issue an identity
to each individual work, each creator and each performer separately.
• Designated authority – ‘The author of an item of metadata should be
securely identified.’ This is necessary to authenticate the metadata and to
provide an audit trail.
• Appropriate access – ‘Everyone requires access to the metadata on which
they depend, and privacy and confidentiality for their own metadata
from those who are not dependent on it.’
The indecs framework forms the basis of the ONIX e-commerce metadata
standard for handling works such as books, sound recordings, graphic arts
and films.
Publishing and the book trade

(Note: Material in this section is based on an interview in November 2016
with Graham Bell, the Director of EDItEUR.)
EDItEUR, the agency that is responsible for the ONIX metadata standard,
takes the view that metadata is there to help solve problems. In the library
context this may be about access and discovery, but for the commercial
publishing industry it is about making the supply chain work. It may be about
making sure a publisher’s books can be discovered by potential purchasers,
whether they are library purchasers or individual consumers in a bookshop.
The transfer of ONIX metadata between agents ensures that wholesalers and
retailers know a book exists so that they can supply it to their customers. In
the book trade metadata can be characterised as all the data that is concerned
with a book from when it is conceived to when it is retailed and everything
in between. It is all the data that is not the book itself. As well as information
about the book, there will be workflow information associated with the
publication, not just bibliographic data. Metadata is used for internal
workflow and for workflow associated with the supply chain as the book is
moved between publishers, wholesalers, distributors and retailers or libraries.
The metadata may also include marketing material to help retailers sell the
book as well as making the book discoverable by individual purchasers.
Because it is expensive to create and maintain metadata, only data that is
strictly necessary to solve a specific business need is created. EDItEUR has
developed technical standards to allow publishers, distributors, wholesalers,
retailers, and the general public to get the metadata easily.
ONIX is based on the indecs model, which was the result of a three-year
project that culminated in 2000. This model has been used to develop a
number of commercial metadata frameworks, including ONIX (for the book
trade), ONIX-PC for periodicals, DDX for the recorded music business and
EIDR for the film and entertainment sector – film, movies, TV. Although on
the surface DDX, ONIX and EIDR appear very different, there are similarities
at a deeper level, because they are based on the same data model.
There is a family of ONIX metadata standards. The classic standard is ONIX
for Books. It is a trade metadata standard for communicating information
about books, e-books and other book-like objects. This might include digital
audio, such as a recording on a CD of someone reading a book.
ONIX-PC is the metadata standard for serials. This metadata is passed
between the serial publishers and aggregators who make bundles of e-
journals available for subscription by academic libraries, for instance.
Bibliographic data exchange – how ONIX works

ONIX is a message that goes between the supplier of the data and the
recipient of the data (publisher and retailer). The publisher and retailer have
their own systems – internal databases containing bibliographic info,
workflow info, and marketing information. The publisher generates an ONIX
record, usually expressed in XML. This metadata is sent to the retailer, who
parses the data and ingests it into their retailer’s database. The databases are
not standardised, but they can both handle ONIX records, allowing for data
to be passed between them as part of an e-commerce transaction. The
publisher’s database will be optimised for the management of the data. They
will be different, but there is a lingua franca for transfer of data between them.
The message is standardised.
Ideally there should be a central aggregator sitting between publishers and
retailers. However in the real world there are a combination of routes, as
illustrated in Figure 9.2 on the next page (Bell, 2016).
In practice there are multiple aggregators and a publisher may use more
than one. Publishers may also deal directly with retailers, and some retailers
(notably Amazon) act as both aggregators and retailers. Some metadata may
pass between aggregators so that a publisher obtains information from
Figure 9.2 ONIX e-commerce transactions
multiple sources. The use of aggregators makes it much cheaper for

publishers and retailers. The use of a common standard reduces the
duplication of effort that would otherwise be necessary to distribute
publications via a variety of outlets. There are some inefficiencies built into
the market. At the time of writing, iBooks, the Apple Inc. system, does not
accept ONIX records. That means that publishers who wish to distribute their
publications via the iBooks platform have to produce duplicate metadata to
Apple’s standards.
Trade in static and moving images

Commercial image libraries have been around for some time and services
such as Getty Images and iStockphoto provide a marketplace for down-
loading digital images for commercial use. Sites that provide a platform for
independent photographers, small businesses and amateur photographers to
sell their images for re-use include iStockphoto, Flickr, Creative Commons
and others. Some providers also allow access to archival images for
educational or even commercial use. In these systems metadata can be seen
as having three roles, which together form the basis of an e-commerce system:
• retrieval and description of images

• management of rights such as re-use of images
• payment transactions to cover the cost of licences.
The IPTC Photo Metadata standard (described in Chapter 4) contains human-

readable metadata which can be embedded in a JPEG or TIFF file. It can also
be held externally in a Digital Asset Management system or in an XMP file.
Music industry metadata

In the 1990s the music industry underwent a major and traumatic
transformation. With the increasing availability of music in digital formats,
first on CDs, then on MP3 players such as iPods, there has been huge demand
for digital music. Increasing storage capacity, improved data compression,
and longer battery life have all created increased demand for digital music.
Enterprising souls have created media-sharing sites where downloaded or
‘ripped’ (i.e. read off a CD and stored on a hard drive) music is made available
free of charge to members, or members are allowed access for a fixed fee. This
has caused a great deal of tension and has arguably undermined the revenue
model of the music industry that has persisted for many years. There are two
aspects to the metadata requirements for digital music.
The first is a retrieval issue. How do you make a music track or an album
discoverable and accessible to your target audience? Apart from the
promotional activity that takes place, it is important that individual
consumers are able to find the track that they want. This may mean offering
retrieval by Artist, Composer, Performer, Band/Orchestra, Date, Location of
the recording, Track name, Album name, Genre, etc. Then there is the
technical data about the compression system used (lossy or non-lossy), the
file format and the resolution of the recording as well as the storage capacity
required to hold it on a portable device.
One of the great benefits of WAV, MP3 and other digital music data formats
is the ability to tag recordings with additional metadata. Apps that use these
formats, such as iTunes and Spotify, can exploit those tags as well as storing
additional information about tracks, such as when they were last played, what
playlists they belong to and images from the album cover. These systems
allow users to tag by genre, other keywords, or simply that they are
favourites. Standards such as MP3 contains a header with technical
information that allows a play-back device to correctly interpret and render
the content to a listener. The main body of the file is a bitstream of the music
itself.
In addition to the embedded metadata for an MP3 or WAV file, there is a

de facto standard, ID3v2, which covers user-oriented aspects of audio
recordings such as Title, Composer, Playlist, Performer, Orchestra or band,
Date of recording, Soloist, and Copyright information. This metadata can be
used by applications for delivery of the recordings to users such as Windows
Media Player, VLC, iTunes or Arpeggio. These tags can also be used by file-
sharing sites and sites that make music tracks available via Creative
Commons licences such as: Mp3.com, Made Loud, Amazon mp3, Sound
Cloud and Jamendo. An MP3 file can be embedded in an ID3 file so that the
metadata tags are handled with the actual music file.
The second aspect of metadata for digital music is about rights
management. This is a commercial issue – about payments and royalties.
Accurate transmission of rights information is vitally important for the
recording industry and crucial for the artists who receive royalties. The
growth of tours by artists who released hits in the 1970s and 80s has been
driven by the need to generate income in the shadow of massively declining
royalties. The music trading sites also use the intellectual property
information associated with MP3 files or included in ID3v2 files, such as artist
names, copyright statements, producers, and links to appropriate websites.
This means that when a track is downloaded via iTunes, Amazon Music or a
similar service, the rights information is passed to the record company which
aggregates this data so that periodic royalty payments can be made to the
artist(s). The aggregated data may be handled by a dedicated rights
management system which keeps track of payments made and payments due
to the artists.
E-government
The Organisation for Economic Co-operation and Development (OECD)
defines e-government as (Field, Muller and Lau, 2003): ‘The use of information
and communication technologies, and particularly the Internet, as a tool to
achieve better government.’ Bhatnagar (2009) suggests that e-government can
be seen as an extension of e-commerce: ‘For those who see it as some form of
extension of e-commerce to the domain of the government, it represents the
use of the Internet to deliver information and services by the government.’ It
encompasses a range of interactions between government and citizens and
between government and businesses. This includes delivery of information,
downloading forms and online form-filling. Examples of e-government in
practice include filing tax returns, registering businesses, applying for
passports and voting. The Federal Government of the USA has identified
metadata elements for description of government digital resources (Central
Information Office, 2016). This schema is primarily for listing data sets in the
Project Open Data directory.
In the UK the move towards e-government prompted the development of
eGMS (e-Government Unit, 2006) the e-government metadata standard, a
Dublin Core application profile. This standard, which is no longer
maintained, and is no longer mandated for use of government websites, was
primarily designed for retrieval of resources on official government websites.
In Australia the Australian Government Locator Service (National Archives
of Australia, 2010) was designed for a similar purpose and also focused on
discoverability of government online resources. Like eGMS it is also an
application profile of Dublin Core and has become a metadata standard for
use within national government. Because much of government information
has a geographic aspect there is some emphasis on spatial metadata in
government metadata standards. Standards such as ISO 19115 (ISO, 2014a)
are coming to the fore and focus primarily on the location of data collections
and services.
The European Commission (2011) has suggested ‘Metadata is an important
asset for eGovernment systems development and as such should be carefully
and professionally managed’. It has been responsible for a number of
initiatives to facilitate exchange of data between public sector organisations
within members states as well as between member states and interoperability
continues to be a priority (Bovalis et al., 2014).
A great deal of e-government’s focus is on Open Data and activity is
currently focused on linked data initiatives, described in Chapters 3 and 13.
This perhaps represents a general move towards greater private sector
participation in the delivery of public services.
Conclusion
E-commerce and e-government have much in common. They are both about
facilitating transactions between suppliers/government and consumers/
citizens. That said, there are some significant differences. E-commerce deals
with transactions between businesses, where there is a need to make the
supply chain work effectively, as exemplified by the book trade system,
ONIX. It also encompasses business–consumer interactions. E-government
tends to focus on electronic transactions between government and citizens or
individuals, although public sector procurement systems are business-to-
business.
CHAPTER 10
Information governance (Purpose 6)
Overview
This chapter considers the ways in which metadata has an impact on information
governance. The first part of the chapter considers the role of metadata in privacy,
freedom of information and legal admissibility of documents. It then goes on to explore
the use of metadata to facilitate regulatory compliance. This demonstrates approaches
to document and information management and to metadata policies which contribute
to the overall ability of organisations to comply with regulatory requirements.
Governance and risk

Information governance is a major preoccupation for many organisations.
Not only is there legislative pressure from freedom of information and
privacy legislation around the world, but there is also a concern about
managing risk. Data loss, malicious access, poor data quality and lack of
interoperability can all have a profound effect on the viability of a business
or integrity of a public service. Reports in the press of data breaches have led
to loss of reputation which can have a real effect on sales. Security breaches
can also lead to direct financial loss as well as lost opportunities and exposure
to further threats.
As well as the generic risks, many corporations operate in regulated sectors
or markets and have to be able to demonstrate compliance. There are
increasing global pressures for multinational enterprises to comply with
national legislation in the markets in which they operate. National
governments and international bodies are co-operating to ensure that
regulations are applied across the board and that potential loopholes are
closed off.
Responses to these pressures by professional bodies such as computing,
information management, document and records management and library
and information services have resulted in new approaches and practices.
These professional bodies also play an important role in setting standards
and providing training for professionals responsible for compliance. This
chapter identifies these areas and looks at ways in which metadata has been
used as a part of information governance. It could be argued that metadata is
an information resource that itself is subject to governance and this is
discussed in Chapter 11.
Authentication
Records management and good governance depends on being able to
demonstrate the authenticity of a record and to provide documentation about
its history and the way it has been managed. This may include details of
transactions that have taken place: who viewed a particular document and
when, what changes were made to the document during its history and the
measures adopted to ensure that unauthorised changes have not taken place.
This provides the basis for legislation on the legal admissibility of electronic
documents and whether they can be used as evidence in legal proceedings.
Provenance and preservation metadata has an important role to play in
authenticating digital documents, as well as providing an audit trail of actions
that have been performed on documents (such as access, amendments, or
deletions). Duranti (1989) puts forward a new role for ‘diplomatics’ (the
authentication of documents – diplomas, certificates or diplomatic docu-
ments). Duranti and Rogers have gone on to develop the idea of diplomatics
adapted for authenticating records stored in the cloud and using metadata to
achieve this:
In digital forensics, the strength of circumstantial digital evidence could be

increased by metadata which record 1) the exact dates and times of any
document sent or received; 2) which computer(s) actually created them; and 3)
which computer(s) received them. Also a chain of legitimate custody (or chain of
evidence, in legal terms) is ground for inferring authenticity and authenticate a
record, and so is a digital chain of custody, that is, the information preserved
about the record and its changes, showing that specific data was in a particular
state at a given date and time. (Duranti and Rogers, 2012, 526–7)
Metadata can assist in compliance by providing an audit trail of who has

INFORMATION GOVERNANCE (PURPOSE 6) 153
accessed personal data or records on individuals and who has updated

personal information. It can also be used to control who has access to which
data. The existence of processes for controlling access is an important strand
in overall information security and accountability and in managing risk.
Information governance
Information governance is an important part of the corporate agenda. Public-
sector and third-sector organisations are also under increasing scrutiny to
demonstrate transparency and to counter perceptions of corruption.
Information governance is a wide term that is taken to mean governance of
information technology, data governance and governance of information held
in documents (whatever their form). Here we concentrate on the last of these
definitions. However, it is important to recognise the overlap between
definitions. Information governance is sometimes seen as part of IT
governance. Definitions are important because to some extent they determine
who is responsible for information governance: lawyers, records managers,
librarians or the IT department. The predominating professional culture will
determine the way in which this issue is handled. The corporate context is
also very important. For instance, regulated industries such as pharma-
ceuticals or financial services have specific reporting requirements that affect
the way in which they handle information.
Although the role of metadata in ensuring information governance is
recognised, there are few practical guides. Blackburn, Smallwood and Earley
(2014) consider some of the questions that arise in information governance
and suggest that metadata may be a way to address some of these questions.
Information governance may in some organisations be closely tied in with
records management and with information security. Both areas are subject to
compliance issues and meeting regulatory standards is one of the major
focuses for information governance activity. The role of metadata in records
management has been discussed in Chapter 7. It looked at the way in which
metadata is used to manage and track records throughout their lifecycle. This
is particularly the case of electronic records and by extension, digital assets.
Information governance may be driven by the management of information
risk, such as the risks associated with data breaches, data loss, disaster
recovery and non-compliance with regulations. In order to get a handle on
this we shall break down information governance into several distinct areas:
information compliance, e-discovery, information risk and sectoral
compliance.
Compliance (freedom of information and data protection)

Freedom of information
Freedom of information (FoI) legislation is in place in more than 100 countries
worldwide, including most of the leading economies. It provides a way of
ensuring that the business of government is transparent and accountable to
the people. However, it means different things in different countries and its
scope varies widely. For instance, in the UK it applies across the public sector
and has been estimated to cover more than 70,000 organisations. Anyone has
the right, regardless of nationality or location, to request access to information
covered by the Act without having to provide a justification or reason. For its
part the public sector in the UK can refuse on the grounds of vexatious or
repetitive requests, cost, or a declared intention to publish the information
according to a published schedule. There are also specific exemptions on the
grounds of national security, commercial confidentiality, current court cases
and the formulation of government policy.
The most obvious role for metadata is to make records discoverable so that
they can easily be retrieved and made available to enquirers. If a publication
scheme has been developed, there will be published categories of information,
which should be included in the metadata of individual documents or records
to allow access via that route. Other captured metadata in a document
management system that may be useful for FoI enquiries include: previous
requests for that information; originating department; file title; author names;
and date of creation. The ability to record this information will depend on
what metadata schema has been used.
Privacy and data protection

The Snowden revelations have generated a great deal of discussion about
information privacy, particularly as it relates to metadata (Greenwald, 2013).
Snowden reported that the US National Security Agency (NSA) had been
requiring telecoms providers such as Verizon to hand over metadata about
telephone calls made between the USA and foreign countries or within the
USA. This was done under the provisions of the Homeland Security Act that
was passed in the aftermath of the September 11th attack on the World Trade
Centre Towers in New York in 2001. Although the NSA did not require the
content of the telephone calls themselves (that is handled by a separate
agency), they required the telecoms provider to systematically gather and
hand over the following:
• details of both parties to the phone call

• time of the call

• duration of call
• when the call was made
• location of the devices (if a cell-phone call)
• unique identifiers.
This has had some effect on legislators and regulators and an impact on trade
relations between major trading blocks such as the USA and the European
Union. Notably, the EU-US Safe Harbour agreement which allowed US
companies to process personal data of EU citizens, was struck down in 2015,
in part because of the systematic and routine gathering of metadata about
phone calls by the US government.
The European Union is characterised by general privacy legislation that
applies across all sectors. The EU General Data Protection Regulation is
principles-based (European Parliament and European Council, 2016). The
European legislation is enforced by national data protection authorities in the
member states. In the USA privacy protection is industry-based, covering
consumers interacting with specific industries such as health (Health
Insurance Portability and Accessibility Act), credit agencies (Fair Credit
Reporting Act) and Federal agencies (Privacy Act of 1974).
Where privacy legislation applies, it is important for information managers
to demonstrate that personal data is handled appropriately. This may mean
codifying the personal data according to how sensitive it is and controlling
who has access to it when and how. For instance, Article 9 of the EU General
Data Protection Regulation prohibits the processing of sensitive personal data
unless strict conditions are met:
Processing of personal data revealing racial or ethnic origin, political opinions,

religious or philosophical beliefs, or trade union membership, and the processing
of genetic data, biometric data for the purpose of uniquely identifying a natural
person, data concerning health or data concerning a natural person’s sex life or
sexual orientation shall be prohibited.
(European Parliament and European Council, 2016)
Personal data is not metadata unless associated with a document of some

kind. However, the management of personal data requires metadata about
that data. Properties of personal data, such as how sensitive it is, are classed
as metadata. Other attributes, such as who input the data, when it was last
updated and whether it has been checked for accuracy, are all metadata
elements that are used to manage personal data about employees, for
instance.
In the context of records management, file plans are a well established way
of determining how documents containing personal data should be handled.
The alternative of allocating security levels to individual documents or classes
of document and then restricting access to those who have the appropriate
level of security clearance can be complicated to administer. Individual
managers may need access to personal details of their staff, but would not
normally have access to personal data of staff in other departments.
Privacy also arises in the case of social media, where individuals create
personal profiles. These are supplemented with behavioural data which is
gathered by the provider and may be made available at varying levels to other
users, and third parties such as online advertisers. Privacy and surveillance
is discussed further in Chapter 14.
E-discovery (legal admissibility)

Companies may be subject to court cases where e-discovery plays a role in
compliance (Tallon, Ramirez and Short, 2013). The use of metadata for
information retrieval is covered in more detail in Chapter 6. Legal
admissibility of electronic documents and digitised images depends on the
accompanying documentation and certification attesting to its authenticity.
Metadata provides a way of recording details of the circumstances of creation
of a document (date of creation, author, editor, etc.) and actions that have
taken place since – an audit trail of who has accessed the document and any
changes that have taken place – what the amendments were and who made
them, and when. As Chapter 2 discussed, many apps (e.g. Microsoft Word)
automatically attach their own metadata to electronic documents when they
are created and this can provide an audit trail for the document as it is drafted
and altered.
There is no way to verify the authenticity of a document without
information about its history and what has happened to it since its creation.
For this, metadata is necessary. The authenticity of information can be
determined by means of physical certificates to indicate that the document
has been checked or that a specific procedure has been followed, or via the
metadata embedded in the resource or held separately in a database.
Information risk, information security and disaster recovery

Data about the sensitivity, source, quality and use of information can form
an important strand of an information governance framework. Metadata itself
can be sensitive and needs to be appropriately managed and protected.
Security of metadata is discussed in Chapter 11. This section considers the
role of metadata in securing information as a part of the management of

information security.
The nature of risk

The use of metadata to manage information risk is predicated upon a good
understanding of what constitutes risk. Many people will recognise a
situation as being risky without a clear understanding of what risk actually
is. A risk is commonly understood to be a threat. A more rigorous approach
defines risk as: an uncertain event with consequences that impact on an area
of activity or interested party. Or simply put, the ‘effect of uncertainty on
objectives’ (ISO, 2009e). There are two dimensions to risk – probability and
impact: How likely is the event to occur? What impact will the event have if
it does occur? There is a tendency to try and quantify risk by expressing it in
financial terms.
Data breaches
Data breaches present a number of challenges to the organisation, including
loss of reputation, loss of customers, regulatory penalties such as fines and
waste of time and effort dealing with the breach. The likelihood of a data
breach may be reduced by an active information security programme in
which metadata plays a part.
Information security
Information security will depend on a number of approaches, including
physical security, hardware, firewalls and communications measures, as well
as procedures and effective management of the data. Using metadata to log
the location of sensitive data, who has access to it and how it is used provides
a mechanism for control and for detecting data breaches. For instance
metadata can be used for forensic analysis of database attacks (Khanuja and
Suratkar, 2014). The previously mentioned use of metadata to establish the
authenticity and provenance of data is also a significant contribution to data
security (Jansen, 2014).
Description of metadata
In the early 2000s some researchers were already developing metadata
vocabularies with the explicit purpose of controlling quality and security of
data. The HIDDEL (Health Information Disclosure, Description and Evaluation
Language) project created a data model describing health websites and health
information providers (Eysenbach et al., 2001). This allows evaluation of
websites and providers but does not go into a great deal of detail about the
individual data elements. De Vries et al. (2014) propose a system of ethical
metadata which provides a context for medical and ethnic data gathered in
the course of malaria research in Africa. Although they do not develop a
specific standard for this type of metadata they describe its potential role and
the benefits:
By way of solution we would propose that at least some information about the
normative context of sample collection and data sharing – what we called ethical
metadata – needs to be taken into account when data sharing decisions are to be
made. This may particularly be the case where research is conducted on
identifiable population groups where stigma or discrimination are of concern.
(de Vries et al., 2014)
Skinner, Han and Chang (2005) describe a concept of Meta Privacy, which
not only encompasses secure metadata, but also uses metadata to manage the
security of a data collection:
An approach that is of relevance to Meta Privacy is the use of meta-information

for privacy protection. Meta privacy tags and metadata can be used for entity
privacy policy preferences representation and enforcement.
(Skinner, Han and Chang, 2005, 58)
Sectoral compliance
There is very little reported research on the role of metadata in compliance.
Kerrigan and Law (2005) report on the development of an engineering
application to extract compliance metadata about environmental regulations
which can then be applied to documents for automatic logic processing of
engineering documentation. Singh and Kumar (2014) consider ways of
complying with data regulations affecting cloud computing. They propose a
four-layer architecture: identification, classification, routing and storage. In
this proposed system data is routed to the appropriate data centres depending
on the type of regulation that applies to it. They talk about metadata
associated with Virtual Appliances (VAP) used to process data so that it ends
up in the appropriate category of data centre.
The REGNET research project in the USA developed a number of tools and
methods ‘to facilitate access, compliance and analysis of government
regulations’ (Law et al., 2014). Among the methods developed, the researchers
use metadata to help with mapping between different terminologies. They

also use metadata about the regulations to retrieve relevant documents, and
relevant sections within those documents. They have also created ‘logic
metadata’ to allow processing by an expert system to represent rules and
concepts from regulations. This is particularly important in reconciling
overlapping regulations and inconsistencies between them.
Conclusion
Metadata contributes to the authentication of documents and data and this is
probably the main way in which metadata is currently used for facilitating
information governance. Information retrieval or (e-discovery) is also
important for access to regulations and this is the other major role for
metadata. Projects such as REGNET have explored ways in which metadata
can play a role in interfaces between regulators and companies through expert
systems, for instance. Using metadata to describe access requirements for data
and to identify sensitive data elements has also been tried. Neither of these
approaches have been widely adopted at the time of writing but may develop
into more-widely available products and services.
PART III
Managing metadata
Part III looks at metadata as a resource to be managed, rather than as a tool
for management that we saw in Part II. Chapter 11 refers back to the metadata
concepts in Part I and identifies some of the issues that arise when developing
and implementing metadata standards, such as quality and security. One way
of addressing the quality issue is to have some control over the way in which
metadata content is created. Chapter 12 considers the ways in which
taxonomies and other controlled vocabularies can be used to improve
metadata quality. Cataloguing rules are also important in this context as are
authority files. Chapter 13 looks at very large collections of data, especially
research data and official data released by public authorities. These require
special consideration because of expansion of linked data and the emphasis
on re-usability of public data. This raises ethical and political issues about the
control and management of information as well as privacy and human rights,
the topic for Chapter 14. This last chapter also peers into the future and
speculates on which professional groups will be responsible for metadata
management and use.
CHAPTER 11
Managing metadata
Overview
This chapter considers the issues surrounding the management of metadata and
describes some of the techniques that are used for metadata management. The project
lifecycle concept is used as the framework for discussion of metadata management. The
management of metadata starts with analysing metadata requirements and moves on
to the development and selection of metadata schemas. There is then a discussion about
encoding metadata and the use of controlled vocabulary before turning to content rules.
Interoperability of metadata schemas focuses on crosswalks and metadata registries.
Quality management covers the use of administrative metadata and reviews issues such
as security of information. The final part of the chapter looks at user education and the
presentation and use of search aids to make metadata more accessible. The chapter
concludes with a view on convergence of management practice for metadata across the
domains.
Metadata is an information resource

Managing metadata, like other aspects of information management, has to
be appropriate to the requirements of its users and fit for purpose. If the
metadata is too detailed, it is costly to maintain. If it is not detailed enough,
the functionality is severely limited. Metadata must be applied in a consistent
way and should be retrievable by those who need access to it. The
management of metadata can be seen as a series of stages, although in most
instances a user will only be concerned with one or two stages in the cycle.
164 PART III MANAGING METADATA
Metadata is an information resource and as such can be described in terms

of a lifecycle with specific activities and processes at each stage:
• Analysing metadata requirements – This will be determined by the main

purpose (or purposes) of the metadata.
• Selecting and developing metadata schemas – Factors such as the nature of
the data being described, the community using it, and pre-existing
conventions and standards need to be taken into account.
• Encoding and maintenance of controlled vocabularies – Many metadata
standards do not specify what encoding is used. Controlled vocabularies
can be developed and maintained using thesaurus techniques.
• Applying metadata – Cataloguing rules can be used to ensure consistency
in the way in which metadata is applied.
• Importing metadata – The choice of source of metadata to import will
depend on factors such as the quality of data available and its
compatibility. Crosswalks and metadata registries provide a means of
mapping between different schemas.
• Quality control – Issues such as security are an important way of
maintaining the integrity and therefore quality of data.
• Search aids and user education – Making users aware of the options
available to them helps them to exploit data sources more effectively.
Workflow and metadata lifecycle

There is no single model of data or metadata lifecycle suited to all applications
and approaches. The aim of this section is to identify some of the models that
have been used and to provide an understanding of some of the principles
that apply to the management of data lifecycles, so that the reader is in a better
position to select and apply operations that are appropriate to their situation.
Greenberg (2009) explores the lifecycle concept as a useful framework for
managing metadata. She proposes to do this in the context of the Dryad
digital repository of data underpinning scientific and medical research
publications. She concludes that, in the absence of any unifying theoretical
framework for metadata, the lifecycle concept is a useful basis for capturing,
creating and processing metadata and the digital materials that it describes.
Previously a ten-stage metadata lifecycle was proposed for digital collections
(Chen, Chen and Lin, 2003) with the following stages:
Group I Requirement assessment and content analysis

1 Acquiring basic metadata needs
2 Assessment of deep metadata needs
MANAGING METADATA 165
3 Review of standards and projects

4 Analysis of elements and standards
Group II System requirement specification
5 Preparation of metadata specification
6 Evaluation of metadata systems
Group III Metadata system
7 Preparation of guidance and best practice
8 Development of metadata system
Group IV Service and evaluation
9 Maintenance of metadata service
10 Evaluation of metadata performance
This ties in with a wider concept of the information lifecycle which has been
expressed in a number of contexts, such as records management, digital
curation and electronic publications. To some extent digital lifecycles can be
applied to metadata, which is itself a type of digital resource. The Digital
Curation Centre’s (2010) digital lifecycle model discussed in Chapter 7 can
be applied to data collections, including metadata collections. The Create or
Receive, Ingest, Preserve, Store, Access and Re-use, and Dispose steps are
most directly applicable to metadata. The Transform step may apply to
conversion of metadata to other formats for export or exchange, or could refer
to the addition and modification of metadata records to reflect the life of a
document or digital resource.
Project approach
An alternative approach is a project lifecycle (Figure 11.1 over the page),
which can be adopted for the development and management of metadata as
described in the previous edition of this book (Haynes, 2004).
In this model the analysis of metadata requirements sets the criteria for
selecting an appropriate scheme or developing a schema or application
profile. The selection may be constrained by issues such as who else is using
this standard, and practical issues of cost of development of a purpose-made
metadata schema. The next stage is to define the vocabulary used in each of
the fields (in database terms, a data dictionary). The metadata is then applied
to items or it may be imported from a third party, which introduces further
issues of cataloguing standards. The quality management processes help to
ensure a consistently indexed resource that is suitable for searching and other
user interactions.
ANALYSING SELECTING AND

METADATA DEVELOPING
REQUIREMENTS METADATA SCHEMAS
SELECT/DEVELOP ENCODING
AND MAINTAIN CONTROLLED
VOCABULARIES
IMPORTING APPLYING
METADATA METADATA
(INTEROPERABILITY) (CONTENT RULES)
QUALITY
ASSURANCE
SEARCH AIDS
AND USER
INTERFACE
Figure 11.1 Stages in the lifecycle of a metadata project
Analysing metadata requirements

A wide choice of metadata schemas is available and new application profiles
and schemas are being developed all the time. One of the challenges of a
manager is to weigh up the relative benefits of developing a schema that suits
a particular application or type of information resource and the use of a
commonly accepted standard that allows for interchange of data. In the
bibliographic world standards such as MARC 21 have long been established
for exchange of bibliographic metadata between applications. This does not
prevent the applications from having their own internal standards – especially
for transactional data such as acquisitions and circulation control. Establishing
the metadata requirements helps managers to navigate through the
challenges of selecting or developing an appropriate metadata schema for the
data resources he or she is responsible for.
Having a clear idea of requirements and a specification will provide a basis
for review of decisions and for reviews of metadata strategy.
It is useful to consider the purpose of the metadata collection. There may
be several reasons for collecting/creating metadata, such as collection
management; resource discovery by learners, preservation, or dissemination
of research. If so, it is likely that one purpose will predominate, although
others will need to be taken into account. For instance, metadata may be used
primarily for retrieval, but description and identification may be other
important requirements.
In practice most metadata schemas have resource description as one of their
purposes but they are usually used for a variety of purposes. For instance,
library standards such as MARC 21, while dealing with resource description,
are also used for information retrieval in library catalogues and resource
management in library management systems. Metadata associated with
records management and preservation is an example of management of
information, but there is often an information retrieval aspect as well.
Metadata associated with ONIX is used for rights management and e-
commerce.
The analysis of requirements has to take into account what is being
described, the systems it needs to interact with, the level of granularity that
has to be supported, the user community, existing standards and the format
of pre-existing metadata. It will also be necessary to take into account the
software environment under which the metadata will operate.
Who will be using the metadata, and how? The profile of users (including
managers and information staff, as well as the eventual audience) is arguably
the most important consideration. What is its overall purpose? This needs to
be taken into account in developing a management approach for a metadata
collection. A requirements checklist might include the following:
• data collection or resource being described

• environment of operation (software)
• availability of existing metadata
• standards
• who will be using the metadata
• resources available to create and maintain the metadata
• need for vocabulary control
• operations carried out on the metadata collections.
Selecting and developing metadata schemas

A sense of the variety of metadata standards available can be had from the
Digital Curation Centre’s (2017) disciplinary metadata resource. This provides
access to a database of metadata standards developed for collections of digital
materials in different subject areas. It also provides links to resources such as
vocabularies, ontologies and tools for handling metadata. Metadata registries
are discussed in more detail later in this chapter. The extent to which a
metadata schema is used and accepted by a community of interest will also
affect the choice of schema. If there is a de facto industry standard or a formal

ISO standard this should be taken into account. In some instances the
standard may be mandatory. However, there may be wider requirements of
the application area itself or specialist requirements that a mandatory
standard may not be sufficiently detailed to address, in which case the
standard becomes one of a number of requirements that should be taken into
account. Working with partners, for instance other members of a cataloguing
union for libraries, or trading partners in the publishing industry, will focus
the choice of schemas to those that are widely adopted within the industry.
For the book trade this could mean the ONIX system or application profiles
based on ONIX. Libraries often use MARC 21 as a common format for
metadata, especially for exchange of data or searching across different
catalogues, and the ability to handle records is a feature of many library
management systems. As with any project, there is a danger of over-
specifying metadata requirements. There are ongoing costs associated with
each piece of metadata and the more detailed and complex, the more
expensive the system will be to set up and to maintain. If extensive indexing
or tagging is required, it may involve human time and effort. In some
instances, such as e-commerce, it may be appropriate to create and maintain
very detailed records, because of the benefits associated with automating
numerous individual transactions. Compromises may also have to be made
between the availability of pre-existing data, which may be beyond
immediate requirements but would be expensive to re-create to another
standard. There are other considerations in selecting and developing
metadata schemas:
• Whatever the primary purpose of the metadata, there will usually be a

requirement to be able to identify individual elements and to retrieve
items described by the metadata. Some metadata schemas are geared to
support search and retrieval capabilities.
• Some data needs to be held in a secure environment, to protect personal
privacy for instance, or needs to be secure against unauthorised access
and interference. This is a particular issue for e-commerce systems. The
security and authentication capabilities built into the schema will affect
the choices available.
• Maintenance of metadata standards – are they stable or rapidly
evolving? If evolving, are they backwards compatible, so that old
metadata is still valid?
• Availability of schema expressed in mark-up languages such as XML or
as RDF.
Importing metadata
Metadata has long been imported from other sources or repositories. Many
libraries import bibliographic records rather than cataloguing new
acquisitions. Cataloguing authorities such as the Library of Congress or the
British Library, or bibliographic service organisations such as OCLC, Neilson,
or Bibliographic Data Services Ltd, sell records to libraries. Importing
metadata requires good selection procedures, quality control and adherence
to a common data standard. Additional work may be needed to clean the data
and to reconfigure it to fit the destination system.
Automatic generation of metadata

The idea of automatic metadata generation has been around for some time.
For instance, Rodriguez, Bollen and Van de Sompel (2009) consider ways of
propagating metadata from existing sources of rich metadata. Metadata can
also be inferred from other metadata expressed in ontologies such as OWL
(Web Ontology Language, see p. 195) or schema.org. In their review of semi-
automatic metadata generation tools Park and Brenza (2015) identify six
methods:
1 metatag extraction
2 content extraction
3 automatic indexing
4 text and data mining
5 extrinsic data auto-generation
6 social tagging.
They conclude that there is a lot of potential for this approach in light of the
large volume of material that needs to be processed and the cost of staff to do
so manually. They accept that some human intervention is necessary, hence
the designation ‘semi-automatic’. Many of the tools that they reviewed use a
combination of these methods. The authors conclude that while the 39 tools
that they reviewed offered many potential benefits, a major barrier to
implementation is the very specific nature of the tools, which were mostly
designed for a very specific domain or data set.
Tools for auto-generation of metadata provide only partial coverage of
metadata elements, which means that human intervention is inevitable.
Automated metadata generation will require considerable investment to
integrate the tools and to make the resulting product more generally
applicable and capable of dealing with a wider variety of sources, document
formats and metadata standards.
Application profiles
Many metadata schemas encourage users to adopt standard metadata
elements that are appropriate to their needs. However, additional data
elements can be created to fulfil specific requirements of the application. It is
also possible to adopt metadata elements from different schemas using a ‘mix
and match’ approach. Application profiles provide a way of re-using existing
metadata standards (or data elements from within those standards) and
facilitate interoperability by the common use of existing standards. Nilsson,
Baker and Johnston (2008) define application profiles in terms of a process to
build functional requirements.
For example, the Singapore Framework for Dublin Core Application
Profiles was developed within the Dublin Core community. It can be used
with other metadata standards, schemas and encoding schemes (Nilsson,
Baker and Johnston, 2008). It has the following components, which are
illustrated in Figure 11.2:
• functional requirements (mandatory)

• domain model (mandatory)
• description set profile (dsp) (mandatory)
• usage guidelines (optional)
• encoding syntax guidelines (optional).
Application profile Usage

Guidelines
annotate
Functional Domain Model Description Set Syntax

built on built on built on
Requirements Profile Guidelines and
Data Formats
built on built on
uses uses
Community Metadata DCMI Abstract DCMI Syntax
built on
Domain Vocabularies Model Guidelines
Models
Domain standards
built on built on
RDF/S built on RDF
Foundation standards
Figure 11.2 Singapore Framework

Interoperability of metadata
What is interoperability?
It is important to have a clear view of what is meant by interoperability, before
explaining the role of metadata in this context. Some definitions focus on the
storage of data in a standard format. A good example can be found in libraries,
where the MARC 21 format is used to exchange bibliographic records between
systems. This does not mean that the library management systems themselves
have to store data internally in MARC 21 format. Indeed, many of these systems
have additional proprietary metadata elements. The internal architecture of the
library management system may make a proprietary data structure more
appropriate. However, the ability to generate output in a standard format and
to import records in an agreed format allows the exchange of data between
systems. For instance, in a relational database system a bibliographic record is
created at the point of querying the system. The different fields comprising that
virtual record are stored in separate tables. Applying a bibliographic standard
that is based on discrete records in a flat file structure may not be easily
translated into a relational system. The indecs initiative defines interoperability
as (Rust and Bide, 2000): ‘enabling information that originates in one context
to be used in another in ways that are as highly automated as possible’. This
definition focuses on the information aspect and the requirement to use
information in different contexts from its origin. It also highlights the
automated nature of transactions.
The above definitions suggest that metadata may be used to facilitate the
exchange of information between systems. However, the data must be capable
of being used by other systems. The implication is that the data is used by
different systems to achieve a common end (such as the successful purchase
of a product). Nilsson, Baker and Johnston (2009) have developed an
interoperability model. It defines four levels of compatibility that can be used
to assess the interoperability of applications with Dublin Core:
Level 1: Shared term definitions

Level 2: Formal semantic interoperability
Level 3: Description set syntactic interoperability
Level 4: Description set profile interoperability
There are two contexts for metadata and interoperability: metadata as a tool
to facilitate exchange of information between interoperating systems, and
interoperability of metadata schemas themselves. Weibel (1998) suggests that
there are three different types of interoperability: semantic, structural and
syntactic interoperability. They are defined as follows:
• semantic interoperability – ‘achieved through agreements about content

description standards’: for example, the agreement between RSC and the
ISBD Review Group for mapping between ISBD and RDA (RDA Steering
Committee, 2015)
• structural interoperability – a data model such as RDF (resource
description framework) that is used for specifying semantic schemas
• syntactic interoperability – an example is XML, which provides a syntax
for expressing metadata. It is about how to mark up and tag data to
enable it to be exchanged and shared with other applications.
In the context of digital libraries, metadata interoperability can be defined in

terms of activities (Arms et al., 2002; Hillmann, Dunsire and Phipps, 2013):
• federation – where metadata from different sources conforms to a

particular standard and is kept up to date. This can be expensive to
implement and so tends to be used where there are significant benefits.
An example of this is interoperability of library catalogues adhering to
Z39.50.
• harvesting – where each participant makes metadata about its collection
available in a simple exchange format. The data is harvested by service
providers. This is good for heterogeneous services. An example is the
OAI – this is considered less expensive for participants and is suitable for
wider participation.
• gathering – used for publicly available metadata, such as that gathered by
search engines on the web. This is the lowest cost option for the data
providers, because no additional effort is required.
• semantic mapping – the repurposing of metadata for use in different
contexts.
The proliferation of metadata standards developed by different, and often

overlapping, communities of interest means that there is a significant danger
of not being able to exchange metadata. With interoperability in mind there
are two sets of opposing pressures on metadata communities. The first is to
simplify the standards as much as possible to ensure that the widest
community can use the standard with minimum effort. This approach has
been adopted by Dublin Core Metadata Initiative. Extensions and refinements
are supported by this approach, while maintaining the integrity of a core set
of data elements. The second pressure is to make the metadata standard
sophisticated enough to encompass the full range of data-handling
requirements that are likely to be required. This is particularly applicable
where the metadata is not only used for resource discovery, but also to
manage the resource and to process transactions connected with the data
entities being described. The ONIX standard and MARC 21 are good
examples of this more comprehensive approach to defining metadata
elements.
Management issues
Use of metadata to enable interoperability brings up a number of
management issues:
• content standards
• suppliers’ interests versus customers’ interests
• cost versus functionality.
Content standards
In order to exchange data there has to be a commonly recognised format for
describing that data, a metadata standard. These standards cover not only
what data is expressed but also how that data is expressed. For instance, an
information resource may have a date field associated with it. An example
would be the date that a recording was made. An agreement on how date is
expressed would be needed between two applications, even if their internal
date representations were different. Frameworks such as Dublin Core suggest
encoding schemes, but they are not mandatory, and it is important that this
is made explicit in the data itself. Using different date conventions is not of
itself a problem, so long as the convention is explicit and there is a way of
converting from one format into another. In a wider context, content
standards need to be agreed between metadata systems so that like is
compared with like and so that the content is interpreted in the appropriate
way.
Suppliers’ interests versus customers’ interests

Some would suggest that it is in the suppliers’ interests to keep their systems
proprietary, so that their customers remain dependent on their systems.
Suppliers are also able to develop unique features and ways of handling data
and transactions that set them apart from their competitors. Users, on the
other hand, want the widest choice and interchangeability of systems. Once
they have opted for a particular system, they want the reassurance of knowing
that they can transfer their data to a new system. They also want to be
confident in selecting a system that allows them to continue to interact with
other applications. The interests of suppliers sometimes conflict with those

of their customers.
Defined metadata standards make it possible for customers to exercise
choice, by either providing a common language for output from the old
system (and for import to the new system), or by defining the data format
used by both systems. The downside of premature adoption of a standard is
that it effectively creates a bias towards one system and stops development
in other areas.
Cost versus functionality

In a paper on the role of standards in interoperability the authors suggested
that increasing the functionality of a standard increases the cost of acceptance
and reduces the number of adopters (Arms et al., 2002). A similar relationship
can be drawn between the functionality of a text mark-up system and the cost
of acceptance.
One of the reasons for the widespread use of Dublin Core as the basis for
application profiles is its simplicity and ease of comprehension. However, its
permissiveness can limit the benefits of exchange of metadata and often does
not deliver sufficient benefit for its use to be justified. This is why many
internet resources are not formally meta-tagged. The benefit of an increased
rate of transaction processing due to interoperable systems can be one factor
that stimulates the development of highly functional metadata standards.
Normalising data
With the proliferation of resource discovery services and collections of
metadata, consistency of metadata has become a major issue. One response
to this is to normalise metadata from different sources. This means that it will
be necessary to use the least specific data available. Although there is a loss
of precision, this is compensated for by the wider range of potential sources
that can be called upon.
A second approach is to require everyone to adhere to the same standard.
This makes sense in communities that have very specific requirements and
where there are benefits to be gained from the additional effort required.
However, this approach is not appropriate for a heterogeneous community
where requirements and purposes may differ quite radically. Importing
metadata from other repositories does raise a number of issues. The iLumina
project (McClelland et al., 2002) identified the following issues:
• Missing elements – There is no control over the quality of external data

and if critical data is missing from a data element this can affect the
interoperability of the resulting service.
• Reconciling values from different vocabularies – If two data providers use
different thesauri for subject terms, this will affect retrieval. Use of
different encoding schemes for structured data such as dates can cause
ambiguities and errors unless there is a way of declaring the encoding
scheme and for translating between different schemes.
• Lack of conventions for using or altering external metadata – It is not always
clear whether permission is given to re-use someone else’s metadata.
• Different field sizes – If the imported data field size is larger than the
maximum for the repository it is being imported to there will be
problems of data integrity and this could cause errors to be reported.
• Inaccurate data – This again relates to the fact that there is no control over
metadata that originates externally.
Some of these issues are addressed in well developed markets for exchange
of metadata and where there are widely accepted standards. For bibliographic
records there is a well established market and reputable suppliers that
provide good-quality data. Even so, there can be variations in the level of
cataloguing undertaken. In other fields it will be necessary to work with some
sample data to establish the feasibility of importing it and to assess its quality
and suitability before undertaking a full-scale import project.
Crosswalks
Reconciling metadata created in different environments is a major challenge
and some effort has been devoted to mapping equivalent metadata elements
between different metadata schemas. These mappings can be displayed as
tables and are known as crosswalks. They can be used within systems to effect
transformations between metadata objects. In the area of bibliographic
standards, BIBFRAME provides a model for bibliographic data that can help
with the creation of crosswalks between schemes. Crosswalks have been
published between Dublin Core and other major metadata schemas such as
MODS. Table 11.1 over the page shows an extract from a Dublin Core to
MODS crosswalk (Library of Congress, 2012).
More complex transformations can be achieved by use of a central
metadata schema for interchange between different schemas. This is similar
to the idea of a key language for translations between many languages. The
advantage of this approach is that there are fewer transformations necessary
to cover the whole range of possibilities.
Table 11.1 Dublin Core to MODS Crosswalk
Dublin Core element MODS element

Title <titleInfo><title>
Creator <name><namePart><role><roleTerm type=text’>
Subject <subject><topic>
<classification>
Description <abstract>
<note>
<tableOfContents>
Publisher <originInfo><publisher>
Contributor <name><namePart>
Date <originInfo><dateIssued>
<originInfo><dateCreated>
<originInfo><dateCaptured>
<originInfo><dateOther>
Type <typeOfResource>
<genre>
Format <physicalDescription><internetMediaType>
<physicalDescription><extent>
<physicalDescription><form>
Identifier <identifier>
<location><url>
Source <relatedItem type=‘original’> + <titleInfo><title>
or <location><url>
Language <language><languageTerm type=‘text’>
<language><languageTerm type=‘code’>
Relation <relatedItem> + <titleInfo><title> or
location><url>
Coverage <subject><temporal>
<subject><geographic>
<subject><hierarchicalGeographic>
<subject><cartographics><coordinates>
Rights <accessCondition>
Figures 11.3 and 11.4 opposite illustrate the concept. This can be expressed
by the formula y = x(x-1), where y is the number of possible connections
(translations) and x is the number of different schemas in operation. This
number rapidly escalates as the number of schemas increases. In a star
configuration, on the other hand, there are x-1 transformations or crosswalks
(i.e. y = x-1). The disadvantage is that except for the key schema in the centre
any crosswalk between two schemas will require two steps rather than one.
It will be necessary to have a crosswalk to the key schema in the middle and
then another crosswalk from the key schema to the destination.
Figure 11.3 shows the possible crosswalks between four schemas using a key
schema compared with direct crosswalks between schemas. There is a total
of eight possible crosswalks between four schemas in the star configuration
(with a key schema in the centre) compared with 12 possible direct
unidirectional connections. As crosswalks are directional, each edge in these
diagrams represents two crosswalks (one in either direction). The star
configuration shows a slight advantage in the number of crosswalks, offset
by the fact that each schema conversion is a two-step process: a crosswalk to
the key schema and then a crosswalk to the destination schema. By contrast
Figure 11.4 demonstrates a significant advantage with 20 possible crosswalks
via a key schema compared with 180 possible direct crosswalks between 10
schemas. The J. Paul Getty Trust adopts a variation of the star configuration
using its own metadata standard, Categories for the Description of Works of
Art (CDWA), as the reference or key schema in the first column of the
crosswalk table. The crosswalk compares data elements from 12 other
Figure 11.3 Possible crosswalks between four schemas
Figure 11.4 Possible crosswalks between ten schemas

metadata standards with those from CDWA (Harpring, 2014).

Initiatives such as the Data Catalog vocabulary (DCAT) are intended to
‘facilitate interoperability of data catalogues published on the web’ (W3C,
2014a). It provides a data model with data elements from Dublin Core Terms
(DCMI Usage Board, 2012), FOAF (Brickley and Miller, 2014) and SKOS
(Miles and Bechhofer, 2009) to describe different aspects of data sets (see
Figure 11.5).
Crosswalks were a popular approach in the late 1990s and early 2000s. A
Figure 11.5 Data Catalog Vocabulary Data Model © 2014 World Wide Web Consortium
(MIT, ERCIM, Keio, Beihang)
number of projects created crosswalks between popular metadata standards.

A crosswalk maps data elements from one metadata schema to another. It is
directional because in many (probably the majority of) instances there is not
a one-to-one mapping of all data elements. In some cases there may not be
an equivalent metadata element (e.g. the PREMIS ‘preservationLevel’ data
elements have no equivalent in Dublin Core). In other instances the
destination may be ambiguous (e.g. Title in DC could map to one or any of
bf:WorkTitle, bf:InstanceTitle, or bf:VariantTitle). Going in the reverse
direction (BIBFRAME to DC) means a loss of information, as the DC.title data
element is less specific than the source BIBFRAME metadata. Some metadata
standards also specify the vocabulary that is used to populate a particular
field. Even if they do not, it may be necessary to translate between
vocabularies. For instance, a bibliographic record from a library that uses

Library of Congress Classification may have to be translated into another
classification coding system used by the destination library such as Dewey
Decimal Classification. So, for example, Yuval Noah Harari’s book History
of Tomorrow is classified in LC as CB428, which breaks down into ‘History
of Civilization. Modern era. 1971-. General Works’. Using Dewey, it is
allocated the code 909.83 – ‘World History’. In other words CB428 in Library
of Congress is equivalent to 909 in Dewey Decimal Classification. There may
also be differences in cataloguing standards which will lead to differences in
the content of fields in the two schemas. So, for instance, the item-by-item
cataloguing approach used for MARC breaks down in a BIBFRAME
environment, where cataloguing may take place at the work or instance level:
bf:WorkTitle, or bf:InstanceTitle.
Metadata registries
Metadata registries provide a resource where metadata definitions and
specifications can be stored and maintained. Many of them conform to the
ISO/IEC 11179 model of metadata registries (ISO/IEC, 2015). They may be
domain-specific or may be maintained by a public authority. A good example
is the METeOR Metadata Online Registry (Australian Institute of Health and
Welfare, 2017). This contains data models used by national health, safety and
welfare agencies and authorities in Australia. Another example is the EPA’s
System of Registers (US Environmental Protection Agency, 2017), which lists
the data standards, data elements and vocabularies used by the EPA. The
Open Metadata Registry (formerly the NSDL Registry) supports metadata
interoperability by providing access to details of 420 vocabularies and 158
element sets (at the time of writing) that have been entered by members of
the registry (Metadata Management Associates, 2017). DataCite metadata
(described in Chapter 13) is accessed via individual research repositories.
Details of the repositories can be viewed via the re3data.org website – the
registry of research data repositories.
Quality considerations
Quality management
The quality management process ensures that the metadata is consistent,
accurate and complete. There are many measures of information quality that
can be applied to metadata. The concept of quality can be applied to the content
of metadata elements as well as to the administrative metadata. The emphasis
tends to be on quality management as a process, which applies throughout the

lifecycle of information rather than being checked at an end point.
Quality of the metadata content

The quality of the metadata content follows some of the same principles as
the quality of information itself. In Chapter 2 we saw how schemas can be
used to determine whether metadata content is in the right form to conform
with the standard. The consistency of the metadata, not only with itself but
also between resources, is important if retrieval is to be consistent and reliable.
For instance, the use of an encoding scheme will help to ensure that the
contents of a particular field or data element are comparable across a resource
or collection of information. Clearly an encoding scheme is in itself not
sufficient to assure the quality of metadata. The skill of the indexer will also
have an effect on the overall quality and therefore effectiveness of the
metadata for retrieval purposes.
Administrative metadata
Administrative metadata shows when the metadata was created or updated
and its origin. Its purpose is to provide a means of managing metadata (as
opposed to the resources described by the metadata). In the early 2000s the
Dublin Core community recognised the need for metadata to describe
metadata, to facilitate interoperability and exchange of metadata. This
culminated in a specialised data element set (Hansen and Andresen, 2003). It
defines the data elements, which are grouped in the following categories:
• metadata for the entire records (such as identifier, scope, language)

• metadata for update and change (based on activity)
• metadata for batch interchange of records (such as filename, technical
format, address of result file).
A useful model of administrative metadata is the A-Core Model (Figure 11.6)

which describes the relationship between what is now known as
administrative metadata and content metadata (Iannella and Campbell, 1999).
describes describes
Resource Content A-Core
Metadata Metadata
Figure 11.6 A-Core Model

Although this original Internet-Draft has not been updated, many of the
ideas in this paper have been carried forward in the Dublin Core DCMI
Administrative Metadata, described earlier. The A-Core Elements are divided
into four components:
1 Who, what, when?

2 Validity dates.
3 Metadata location.
4 Rights ownership.
The provenance of metadata is an important aspect of quality. The

provenance of the metadata is intended to answer the following questions:
Who created the metadata? When was it created or amended? What were the
circumstances that led to the creation of the metadata? Chapter 8 discussed
metadata as an information object, for which the provenance can be recorded
as part of preservation management.
Metadata security
Security is a major consideration in an interoperable environment. A useful
analysis by the New Zealand government suggests that security is a key issue
(New Zealand E-Government Unit, 2001):
• to guarantee integrity of data

• to make data manageable by grouping by agency or user
• to prevent alteration of data by other agencies
• to control access to system functions.
At a basic level a security strategy for metadata will need to ensure that the
metadata maintains its physical integrity, by being stored securely on a
system with regular back-ups. The storage medium will be subject to all the
same considerations that would apply to any kind of electronic data:
robustness of the storage medium, corruption of data by decay of the
medium, storage conditions for the medium, durability of the medium,
technology used to read the medium. A strategy for back-up and migration
of the metadata will go some way towards addressing these concerns.
It will be necessary to restrict editorial access to metadata to authorised
personnel. The access is usually controlled by the operating system. At a crude
level it can be used to allow only certain people access to the metadata
management application. For example, different levels of access might
include:
• Read – The user can view metadata and print it off. In some cases this
will extend to the issue of whether or not the user even knows that the
record exists.
• Create – The user can create new metadata records.
• Edit – The user can amend or edit existing records – normally the date of
any changes and the name of the person making the change is recorded.
• Delete – The user can remove a record from the system – although an
audit trail should indicate that this has been done.
The levels of access can be fine-tuned so that individual records or even data
elements may have their own security levels. Users are then assigned a
security authorisation that allows them appropriate access to records or data
elements. These measures depend on the ability to identify individual users
and to control their access to the system. Most commonly a user’s identity
and password provides a basic level of security. More sophisticated systems
may require some kind of physical verification such as a key. This may be an
electronic key such as a swipe card, or could be based on a physical attribute
of the user such as a finger-print or iris image.
Another aspect of security is data privacy. If the metadata is being stored
as a back-up on a removable medium for instance, or being transmitted from
one location to another over the internet, it may be necessary to encrypt it.
There has been a lot of discussion about the balance between national security
and privacy in light of the extensive use of communications metadata by
bodies such as the US National Security Agency (Solove, 2011; Morrison, 2014;
Greenwald, 2013). Metadata itself needs to be kept securely as well as playing
a part in the security of the data that it describes. Skinner, Han and Chang
(2005) introduced the concept of ‘meta privacy’ to deal with the issue of
protecting metadata, as distinct from using metadata to protect data. They
talk about meta privacy in terms of benefits and risks associated with secure
metadata. They advocate the use of privacy tags that are attached to metadata
elements, which govern access to the contents of that data element.
Conclusion
Metadata is an information resource that needs to be managed. A lifecycle
approach can be adopted to handle metadata throughout its life. An alternative
approach is to view metadata creation as a project and to apply project
management principles to it. This means analysing metadata requirements,
selecting and developing a schema and then importing metadata. Application
profiles represent another option for development of suitable metadata
standards. The Singapore Framework for the development of application
profiles encompasses the project approach to metadata management.

Interoperability of metadata is necessary for development of linked data
applications. Crosswalks provide powerful ways of converting metadata from
one standard to another. Where there are crosswalks between several
standards, it may be more efficient to establish a key schema through which
all metadata standards are translated. It is just as important to manage the
metadata quality as it is to manage the accuracy and consistency of the
documents it describes. Security is required to ensure the integrity of the
metadata and to protect personal privacy.
CHAPTER 12
Taxonomies and encoding schemes
Overview
This chapter is all about the content of metadata elements. Permissive standards such
as Dublin Core describe what each field or data element is for, but do not specify how
the content of that data element is generated. For instance, the ‘dc:creator’ data element
might contain the name of an organisation as it is known to the website manager, it might
be taken from an authority file, or it may be created according to a set of cataloguing
rules such as AACR2. This chapter is about the techniques or mechanisms that are used
to manage and control the content of individual data elements. This is important for
consistency, quality of retrieval and efficiency of operation. Controlled vocabularies,
authorities and cataloguing rules all come under the heading of encoding schemes. A
more detailed treatment of cataloguing can be found in Welsh and Batley (2012). The
development and use of controlled vocabularies are covered in classic works such as
Aitchison, Gilchrist and Bawden (2000) and more recently in Broughton (2006). Lambe
(2007) and Broughton (2015) deal with aspects of classification and taxonomies – also
important sources of terms for metadata elements.
Role of taxonomies in metadata

Hedden (2016) recognises the convergence of metadata and taxonomies and
the richness of software applications to handle taxonomies and ontologies. It
is key to managing the content of metadata elements. Lambe (2007) sees
metadata as one way of instantiating a taxonomy, the other being a thesaurus.
Increasingly taxonomies and controlled vocabularies are being incorporated
into document and information management products and services such as
Sharepoint 2013. White (2016) makes a strong case for the use of taxonomies
and controlled vocabularies in an enterprise search environment. Bloggers
such as Earley (2017) and the Metadata Research Center (Drexel University,
2017) have also contributed to discussion about metadata and taxonomies.
Encoding and maintenance of controlled vocabularies

One of the strengths of metadata schemas such as Dublin Core is that it
provides a means of comparing the content of data elements for different
resources. Each element has a defined meaning, so that there is a semantic
relationship. This means, for instance, that the Creator data element will
contain information about the person, group or organisation responsible for
creating the resource. This provides a mechanism for implementing the
semantic web, where like can be compared with like. However, unless there
is some agreement about how that data is expressed, the benefits are limited.
This can relate to fundamental attributes of data such as what language it is
expressed in. The following marked-up text indicates that the content of the
data elements is in English (of the British variety):
<meta name=‘DC.Title’ xml:lang=‘en-gb’ content=‘Home ownership’ />

<meta name=‘DC.Creator’ xml:lang=‘en-gb’ content=‘Shelter, England’ />
<meta name=‘DC.Subject’ xml:lang=‘en-gb’ scheme=‘LAMS-CCS’
content=‘Home ownership’ />
In the case of subject retrieval an indexer may have to select terms from a
controlled vocabulary such as a thesaurus or from classes in a classification
scheme or taxonomy. This is especially important when dealing with a
structured collection of material where it is necessary to reliably and
consistently retrieve relevant material according to search criteria established
at the point of need. Using a controlled vocabulary ensures more consistent
retrieval. This limits the searcher to a preferred term choice rather than having
to think of what synonyms might describe the concept being searched for. In
records management systems a file plan provides a similar mechanism,
allowing users to select files according to a designated category which may
be subject-based or based on a functional analysis. The selection of terms or
categories can be presented as drop-down lists, as searchable databases, or
as navigable networks of terms. Many specialist organisations have
developed their own thesauri tailored to their needs. This approach has also
extended to EDRM (electric document and records management) systems,
where subject retrieval is a key consideration. A thesaurus allows a range of
relationships between terms to be included. A full treatment of thesaurus
TAXONOMIES AND ENCODING SCHEMES 187
development can be found in Thesaurus Construction and Use (Aitchison,

Gilchrist and Bawden, 2000) and Essential Thesaurus Construction (Broughton,
2006). Some aspects of thesaurus construction and use are also covered in ISO
25964-1 (ISO, 2011). Lists of specialist thesauri and taxonomies can provide a
good resource for identifying a controlled vocabulary to adopt for specific
metadata fields, or they can provide a source of terms for a tailored
vocabulary (Taxonomy Warehouse, 2017; Basel University Library, 2017;
University of Toronto Library, 2015). If a controlled vocabulary is necessary
(sometimes it is not; for instance in free-text titles), there are three options:
1 Adopt external controlled vocabulary – If a similar organisation has

developed its own thesaurus, or there is a thesaurus covering your
organisation’s areas of activity, then this may be a cost-effective
approach. It saves the effort of generating your own terminology and has
the advantage of being in line with at least one other organisation. The
disadvantage is that you have no control over the development of the
thesaurus and incorporation of new terms.
2 Select pre-existing standards – There are standards for encoding particular
types of data such as dates, ISO 8601 (ISO, 2004c) and languages, ISO 639-
1 (ISO, 2002). By their nature they are widely adopted, as they tend to
reflect a consensus across a wide range of users. This approach tends to
work for very specific and clearly delimited areas.
3 Create controlled vocabulary – This is the most ambitious and expensive
option, as it requires an analysis of the subject coverage and functions of
the organisation, and considerable effort to compile. It has the advantage
of being tailored to the needs of your organisation and of being under
your control – so you decide what new terms are added or which are the
preferred terms.
There are many tools for developing and maintaining controlled vocabularies
such as thesauri and taxonomies. A good starting point is the Thesaurus
Software Directory (originally on the WillPower website and now maintained
by Taxobank) and the links to online resources mentioned in Heather
Hedden’s book The Accidental Taxonomist (Will and TaxoBank, 2013; Hedden,
2016). Other lists have been produced periodically but have not been kept up
to date. There are also professional groups and discussion lists that have an
interest in taxonomies or classification schemes, such as the American Society
for Information Science and Technology (ASIS&T, 2017), the Special Libraries
Association Taxonomy Division (SLA, 2017) and the International Society for
Knowledge Organization (ISKO, 2017).
The ISO 24617 series of standards provides guidance on different types of

semantic annotation, including time and events, dialogue acts, semantic roles,
discourse structure, special information and semantic relations in discourse
(ISO, 2016b). It supports the development of interoperable resources and is the
precursor to representation of information using a specific language such as
XML. This framework is a modelling system for semantic elements of resources
and can therefore be used for developing rules for encoding schemes.
Thesauri and taxonomies

Thesauri and controlled vocabularies present a number of opportunities to
enhance the quality of retrieval from a document collection:
• If the organisation has invested effort in defining a standard vocabulary

or set of terminology, it is possible to enrich metadata with the
‘preferred’ terms from the thesaurus, and to configure the search engine
to look for the appropriate metadata.
• Indexing can be automated so that documents are processed to identify
potential descriptors. The preferred term is associated with the
document as metadata when a recognised synonym is found in the text
of the document. This ensures that a consistent term is used to describe a
concept regardless of the actual words used. This approach can be
enhanced by providing users with a drop-down list or a navigable
hierarchy to find suitable ‘preferred’ search terms.
• A taxonomy can be used to classify and organise documents in the
collection. The taxonomy may be applied manually or automatically
based on recognition of words in the text. The user then has access to the
resources by means of a map or drop-down lists of categories under
which the documents fall.
Advances in automatic indexing and classification of resources to enhance

the metadata have highlighted the need for a framework for evaluating the
performance of auto-indexing systems (Golub et al., 2016).
Synonym rings
Many search engines support synonym rings. These can be created from a
simplified thesaurus, which associates terms that are synonyms or quasi-
synonyms (words which have the same meaning or similar or related
meanings). For instance, the following terms could be associated with one
another in a synonym ring:
Thesaurus relationships
Thesaurus relationships are defined in ISO 25964-1 (ISO, 2011) and an excellent and
detailed description of them is given in Aitchison, Gilchrist and Bawden (2000). A term
may be associated with other terms, defined by relationships. If we take ‘Bacteria’ as
our lead in term (Haynes, Huckle and Elliot, 2002) the following relationships can be
defined:
Bacteria
BT: Microorganisms
BT: Pathogens
NT: E coli
NT: Legionella
NT: Listeria
NT: Salmonella
BT – Broader Term. This is a more general term and is higher up the hierarchy. The
thesaurus may have several levels of hierarchy – which can provide a useful navigation
tool. In this example ‘Bacteria’ has two Broader Terms, ‘Microorganisms’ and
‘Pathogens’.
NT – Narrower Term. A more specific term, lower down the hierarchy. A term may have
more than one narrower term. The inverse relationship is a broader term. The
narrower terms of ‘Bacteria’ are: ‘E Coli’, ‘Legionella’, ‘Listeria’ and ‘Salmonella’.
The next relationship is illustrated by the example of ‘Addiction’ from the HSE
Thesaurus (Haynes, Huckle and Elliot, 2002).
Addiction
BT: Psychiatric disorders
RT: Alcohol abuse
RT: Drug abuse
RT: Smoking
RT: Substance abuse
RT – related term. This is for terms that are associated with the term in question. This
is a useful way of broadening the search or providing a route to alternative search
terms (or indexing terms). This feature can be particularly helpful for generating drop-
down lists of alternative search terms. In this example entering ‘Addiction’ would
produce a drop-down list of alternative search terms including ‘Alcohol abuse’, ‘Drug
abuse’, ‘Smoking’ and ‘Substance abuse’.
The final use and use for relationship is illustrated with the following example:
Personnel managers
UF: Human resources managers
UF: Industrial relations managers
UF: Training managers
BT: Functional managers
USE – preferred term. This points to the preferred term. A thesaurus represents a
‘controlled vocabulary’ to ensure consistency of indexing (and retrieval). The entry for
Training managers in the thesaurus would have the USE ‘Personnel managers’ as its
entry.
UF – Use For, i.e. non preferred term. This points to synonyms of a preferred term. In
this example ‘Personnel managers’ is the preferred term and the UF relationships point
to the non-preferred terms that would be synonyms.
Personnel managers, Human resources managers, Industrial relations

managers, Training managers.
Another example of a synonym ring groups together related terms from a

thesaurus:
Addiction, Alcohol abuse, Drug abuse, Smoking, Substance abuse.
In a structured thesaurus, these terms would be related to one another with

the following relationships:
USE
USE FOR
RT (related term)
NT (narrower term)
A search for any one of these terms would retrieve all the terms in the
synonym ring. This improves recall at the expense of precision. Precision can
be improved by being more selective in the relationships included in the
synonym ring – for example by limiting the synonym ring to true synonyms,
defined by the USE and USE FOR relationships. An alternative approach is
to be more inclusive (by using quasi-synonyms defined by RT and NT
relationships as well as USE and USE FOR), but to generate drop-down lists
in response to queries and allowing users to explicitly select related terms.
Role of controlled vocabularies

A thesaurus is a representation of knowledge, the relationships between terms
reflecting assumptions about the nature of a subject or discipline.
The terms ‘taxonomy’, ‘thesaurus’ and ‘classification’ are sometimes used
interchangeably. However, it is useful to make distinctions between these
terms. A thesaurus is normally made up of terms (words or phrases) which
represent single concepts. These terms are classified for ease of management
and may be displayed as a hierarchy or as an alphabetic list. Taxonomies and
classification systems are ways of organising or grouping entities, whether
they be species of insects, document collections, web pages on a domain or
ideas about a subject. Each category in a taxonomy may be a complex concept
made up of simpler concepts, or it may be a class of objects used for
discriminating between objects.
In indexing theory there is a distinction made between pre-coordinate and
post-coordinate indexes. A pre-coordinate index puts simple terms together
to produce a category or term that can be used for searching. For example,
an index in a recipe book may be organised by dishes, e.g. Carrot cake, which
would be a pre-coordinated index term. In contrast, a searchable database of
recipes may have an index term ‘Carrots’ and a separate term for ‘Cakes’. This
is a post-coordinate system, because the terms are put together (or
coordinated) at the search stage rather than the indexing stage.
Content rules – authority files

When applying metadata, consistency is important. The metadata may be
applied manually or may be partially automated (based on recognition of
synonyms in the text) and it may be embedded in the resource or kept in a
separate database or repository. Not all elements can be populated from a
limited, controlled list of terms. For instance, personal names, company
names and addresses are all variable, but still need to be quoted consistently.
A set of cataloguing rules or conventions can help to ensure that a particular
name appears in a consistent form and that users (knowing the cataloguing
conventions) can easily find appropriate entries. Of course libraries use
authority lists for standard forms of names or construct them using
cataloguing rules such as RDA or ISAD(G).
It is easy to see the difficulties that arise if there is no established convention
for expressing names. For example, Jane E. Smith could be expressed as Dr
Smith; Smith, Jane; Smith, J.; Smith, Jane E. etc. Each alternative will affect
retrieval. The authority file extract from the Library of Congress (see Figure
12.1 on the next page) also would have difficulty distinguishing between
authors where Smith, J. is used. Even being as specific as ‘Jane E Smith’
produces some ambiguous results.
If the data is to be processed automatically, the parts of the name and the
order in which they appear can be critical. Comparison of two items becomes
difficult if different conventions are used to generate identifying metadata
such as ‘author name’.
As well as being a source of metadata content (e.g. name authorities for
authors), authority files are themselves structured and defined by metadata
standards. The Functional Requirements for Authority Data (FRAD) provides
a data model for authorities such as the Name authority file maintained by
the Library of Congress (IFLA, 2013). Other standards such as Resource
Description and Access (RDA) also incorporate this model. Figure 12.2 shows
the relationship between bibliographic entities (such as books) which are
known by identifiers such as ISBNs or by Names such as author names. This
forms the basis for an author index (controlled access point) or a Name
authority file. This is governed by Rules (such as RDA) which are applied by
an Agency, such as the Library of Congress.
Figure 12.1 Extract from an authority file from the Library of Congress
>> known by
Bibliographic
Entities
>>
Names and/or >>
Identifiers
basis for
>>
Controlled
is based on Access Points
>>
is governed by
>
governed
>> Rules
are applied by
>>
applied Agency
is created / modified by
>>
creates / modifies
Figure 12.2 Conceptual model for authority data (based on IFLA, 2013)
Creating metadata values is a key management issue. Some systems, such

as EDRM systems, may depend on users creating some metadata, combined
with metadata derived automatically from the system. For instance, when a
new electronic file or record is created, the responsible person may have to
select a category for that file from the file plan or classification scheme. That
person will probably also create a title and may add keywords to enrich the
indexing of the file so that others can find it in the future. Some information,
such as the date the file was opened; the name of the person opening it and
the department, would be generated automatically by the system.
Some systems allow for automatic generation of metadata. So, for instance,
the text can be processed to select appropriate keywords from a controlled
vocabulary, which is presented to searchers when they want to identify and
retrieve a file. Figure 12.3 illustrates the way in which natural language terms
are associated with controlled terms from a thesaurus. These can be used in
a search interface to retrieve all synonyms or they can be used to enrich the
metadata associated with the document to ensure reliable retrieval.
The Tokay d’Alsace

grape variety makes
robust white wine Tokay d’Alsace USE Pinot Gris
that is an excellent porgy USE sea bream
accompaniment to
porgy fish
Metadata
Keword: pinot gris
Keyword: sea bream
Figure 12.3 Use of terms from a thesaurus
Although indexing can be done by the author of a document and index

terms can also be generated automatically by some systems, for highly
structured systems such as library catalogues, professional indexing may be
preferred for a better-quality, more consistent result. For certain technical
information such as format data associated with images, for instance,
automatic capture of embedded metadata is widely used. Manual indexing
is expensive, and it is not always possible to make this level of investment.
Alternatives such as latent semantic indexing and automatic analysis of text
to construct indexes are offered by some search engines as an alternative to
human intervention.
Intelligent searching systems that ‘learn’ users’ information requirements

provide another route to accurate and comprehensive retrieval of relevant
material. This is an alternative to the use of metadata to describe a resource
and works well where individual needs are constant and the sources being
searched are very diverse (as on the internet).
Digital resources with embedded metadata facilitate local management as
well as transfer of those resources between different repositories. Networks
of repositories working in co-operation can develop sophisticated back-up,
migration and storage strategies to ensure that secure copies of the images
are maintained while keeping storage requirements to a minimum.
Ontologies
There is some discussion about the difference between ontologies and other
types of classification system. ‘An ontology is a set of precise descriptive
statements about some part of the world (usually referred to as the domain
of interest or the subject matter of the ontology)’ (W3C, 2012). Gruber talks
about ontologies in the following terms:
In the context of database systems, ontology can be viewed as a level of

abstraction of data models, analogous to hierarchical and relational models, but
intended for modeling knowledge about individuals, their attributes, and their
relationships to other individuals. (Gruber, 2009)
In other words, an ontology is a way of modelling reality and as such is a

knowledge representation.
McGuinness (2002) describes three universal properties of all ontologies:
(i) a finite controlled vocabulary, (ii) unambiguous interpretation of classes
and term relationships; and (iii) strict hierarchical subclass relationships
between classes. In effect an ontology takes the form of a specialist type of
classification system or taxonomy. McGuinness goes on to describe uses of
ontologies and in particular their application in a web environment. These
range from a source of controlled vocabulary, to a way of organising content,
to improvement of navigation, browsing and retrieval capabilities. More
complex ontologies can be used for modelling data. Lambe (2007, 238) also
talks about ontologies in terms of concepts and the relationship between
them:
Concept A – Relationship – Concept B

This means that they can be expressed as RDF triples. They are designed
for processing on computers and allow for the creation of new relationships
based on existing relationships and inferences that can be drawn from them.
Corcho, Poveda-Villalón and Gómez-Pérez (2015) talk about ‘lightweight’ and
‘heavyweight’ ontologies. Ontologies contain concepts and the relationships
between them. In these terms a thesaurus would count as a lightweight
ontology. However, a heavyweight ontology would contain more complex
relationships than those typically found in a thesaurus. They would also
contain axioms that enable applications to define new relationships. This has
proven useful in specific areas of research such as genetics.
OWL (Web Ontology Language)

OWL (Web Ontology Language) was developed specifically to handle
concepts and the relationships between them (OWL Working Group, 2012).
OWL is described in the following terms:
OWL is a computational logic-based language such that knowledge expressed in

OWL can be reasoned with by computer programs either to verify the
consistency of that knowledge or to make implicit knowledge explicit. OWL
documents, known as ontologies, can be published in the World Wide Web and
may refer to or be referred from other OWL ontologies.
(OWL Working Group, 2012)
It is now in its second release, OWL 2, and has the following features (W3C,
2012):
• axioms – the basic statements that an OWL ontology expresses

• entities – elements used to refer to real-world objects
• expressions – combinations of entities to form complex descriptions from
basic ones.
An axiom might be: ‘books have a publication date’.

An entity might be: ‘The Hitch-hiker’s Guide to the Galaxy is a book’.
An expression might be: ‘therefore the book The Hitch-hiker’s Guide to the
Galaxy has a publication date’.
Some general-purpose ontologies have been developed, such as FOAF (Friend

of a Friend) for representing connections in social networks and on the web
more generally (Brickley and Miller, 2014). Another example is SKOS (Simple
Knowledge Organization System), which describes the structure and content
of topics, controlled vocabularies, classification schemes, taxonomies and

folksonomies (Miles and Bechhofer, 2009). SKOS is an OWL ontology and SKOS
vocabularies are instances of the ontology. Examples of vocabularies (data sets)
listed on the SKOS website include: Unesco thesaurus, Agrovoc Agricultural
thesaurus, VIAF Person Authorities (from the Virtual International Authority
File), and Language Codes based on ISO639 (W3C, 2015).
Schema.org
Schema.org is an ontology system based on semantic web technology with
controlled vocabularies for digital objects such as web pages, digital sound
recordings, images and electronic publications (Sponsors of Schema.org, 2017).
Schema.org metadata can be expressed in RDFa, Microdata and JSON-LD. It is
widely used by search engines and is the result of a collaborative effort by
Google, Microsoft, Yahoo! and Yandex. The vocabularies have been developed
by an open community process and are continually developing. Schema.org
can be extended either as a hosted extension (reviewed and managed by
schema.org), or as an external extension (managed by other groups). In effect
schema.org allows for tagging of content using a common vocabulary for
entities, relationships and actions. Schema.org contains 589 types, 860
properties and 114 enumeration values. Commonly used item types include:
• Creative works: CreativeWork, Book, Movie, MusicRecording, Recipe,

TVSeries . . .
• Embedded non-text objects: AudioObject, ImageObject, VideoObject
• Event
• Organisation
• Person
• Place, LocalBusiness, Restaurant . . .
• Product, Offer, AggregateOffer
• Review, AggregateRating.
The schema.org website uses the film Avatar to demonstrate how the mark-
up works (Sponsors of Schema.org, 2017). It declares Avatar to be item type
‘Movie’ (expressed using Microdata tags in HTML 5):
<div itemscope itemtype=‘http://schema.org/Movie’>

<h1>Avatar</h1>
<span>Director: James Cameron (born August 16, 1954)</span>
<span>Science fiction</span>
<a href=‘../movies/avatar-theatrical-trailer.html’>Trailer</a>
</div>
Schema.org can be used to specify the properties associated with the item:
<div itemscope itemtype =‘http://schema.org/Movie’>

<h1 itemprop=‘name’>Avatar</h1>
<span>Director: <span itemprop=‘director’>James Cameron</span>
(born August 16, 1954)</span>
<span itemprop=‘genre’>Science fiction</span>
<a href=‘../movies/avatar-theatrical-trailer.html’
itemprop=‘trailer’>Trailer</a>
</div>
Schema.org vocabularies are used by the main search engines to improve the
relevance of search result rankings. Tagged content using Schema.org
vocabularies and mark-up enables users to search in context and obtain more
precise search results. For instance, Google uses content that is tagged using
the Schema.org vocabularies (as well as other vocabularies) to populate its
Knowledge Graph database (Singhal, 2012). It provides factual answers to
queries without having to go to the external websites themselves.
For instance, a museum could tag its home page as a museum, which is
listed as Thing-Place in Schema.org. A search on google.co.uk yields at the
top of the search results the result in Figure 12.4.
Figure 12.4 Google Knowledge Graph results

Clicking on ‘The British Museum’ then shows structured data derived from
the schema.org metadata for that museum, without having to go to the
museum’s website (Figure 12.5).
Figure 12.5 Structured data in Google about the British Museum

Social tagging and folksonomies

Social media applications encourage active participation by users in a number
of ways: creating content; uploading images; sharing reports on activity;
connecting to other people; commenting on posts by other people; and pro-
viding emotional support through the ‘like’ feature. In many social networking
environments, users are also left to their own devices when it comes to tagging
content or providing descriptions. This has led, especially in the image
exchange communities such as Instagram and Flickr, to the emergence of ‘folk-
sonomies’. The consolidated collective effort of users is viewed as an enterprise
to make content accessible. This is particularly the case for images with no text
associated with them. This means that free-text (or natural language) searching
is not available to users. The problem is that there is minimal co-ordination –
users do have access to lists of tags used by previous users, but the resulting
vocabulary is so large and uncontrolled that it is difficult in practice to navigate
effectively to find the appropriate term(s) or subject headings.
Self-indexing is also a feature of institutional repositories, where many
institutions require authors to index papers that they have submitted to the
repository. Rarely are they skilled indexers and very often there is very little
interest or motivation to spend much time on what might be regarded as a
chore. Some suggestions have been made for the development of automated
indexing approaches combined with a folksonomy approach. However, there
has to be a tested framework for evaluating the performance of such systems
against the alternatives (controlled language or self-tagging exclusively) –
Matthews et al. (2010); Golub et al. (2016).
Measures of tag weighting

Syn and Spring (2013) have created two measures for tagging that are intended to help
managers of metadata collections to assess the quality of social tagging:
Annotation Dominance (AD). This considers the number of times a particular tag is
applied to a resource relative to the total number of users who have used tags. This is
a measure of the level of consensus or agreement about the application of a particular
tag to a specific resource. For example, a photograph on a social media site might have
the tag ‘sunset’. The AD measure would be an indication of the proportion of taggers
that have used the tag ‘sunset’ for that particular picture. The value of AD will be
between 0 and 1.
(Count (TAi ,Rj )
AD =
Count (U,Rj )
Here, the numerator, Count(TAi, Rj), indicates the number of tag sets that contains a
tag Ai (TAi) assigned to a resource Rj, and the denominator, Count(U, Rj), is the
number of all users who bookmarked a resource Rj with a tag set. Therefore, AD is a
measure of how much a tag is agreed by users to represent a given resource.
(Syn and Spring, 2013, 969)
Cross Resources Annotation Discrimination (CRAD) This is a measure of the usefulness of

a tag for discriminating between categories of document. If a tag is used extensively
(an extreme case would be applied to every document) it is not useful for
discriminating. If a tag is used for only one document, it is also of little use for
categorising documents. Where a tag is used more than once (i.e. Count (U, Ai) > 1):
log
(Count (R, A ))
Count (R)
i
CRAD =
log(Count (R))
When the CRAD and AD values are multiplied together they provide a measure of the
degree to which a tag can act ‘as semantic and classificatory metadata terms’ (Syn and
Spring, 2013, 971).
Stock and Stock (2013, 611–14) characterise folksonomies as ‘knowledge

organisation without rules’. They suggest five ways of improving the quality
of social tags, building on previous work on ‘tag gardening’ (Stock and Stock,
2013, 621–8; Peters and Weller, 2008):
• Weeding to remove bad tags such as spam, spelling errors and variations
on standard spelling (e.g. standardising on the US spelling ‘sulfur’ or the
British spelling ‘sulphur’ and removing errors such as ‘sulpur’ and
‘slufur’. Another example would be deciding whether the preferred term
is ‘epinephrine’ or ‘adrenaline’ and automatically re-indexing the non-
preferred terms.
• Seeding with ‘previously allocated tags’ to ensure that frequently used
tags are considered for each new document. For example, WordPress
lists previously used tags when a blog author is making a new entry. The
size of the tag text indicates which tags have been used most frequently.
• Vocabulary control to conflate synonyms and to disambiguate
homographs. For instance, on Mendeley, books about ‘information
retrieval’ may have the following tags: ‘IR’; ‘info retrieval’; and
‘information retrieval’. Standardising on one of these is desirable. An
image on iStock with the tag ‘plant’ could refer to an industrial facility or
a living thing and needs to be disambiguated to make it clear which
meaning is intended.
• Fertilizing by offering semantically related items to a user. A search on
‘plant’ also yields results for ‘tree’, ‘flower’ and ‘leaf’ on iStock.
• Harvesting to identify and use the most popular tags or ‘power tags’ to
index documents. For instance, catalogue entries for books on
LibraryThing display the tags that have been assigned by other users.
The size of the tag text indicates which ones are most popular and thus
direct other users to them.
Guy and Tonkin (2006) provide a useful overview of the origin of

folksonomies as well as making suggestions to improve the quality of tagging.
Conclusion
In order to be useful for retrieval, management or interoperation, not only is
there a need for an agreed metadata standard, but also the content of the
metadata elements needs to be managed. Encoding schemes such as
controlled vocabularies, authority lists and cataloguing rules are all well
established methods of achieving this. However, in the era of linked data a
more sophisticated approach is required to allow for more complex
relationships between concepts and to facilitate the processing of data
elements to create new data. The development of languages such as OWL and
schema.org provide a mechanism for this and this development has resulted
in the creation of general-purpose ontologies such as FOAF, SKOS and
schema.org. A large number of specialist ontologies have also been developed
in specific areas such as genetics.
The establishment of the semantic web and the growth of social networks
that allow interaction between users and systems have led to the proliferation
of social tagging and the growth of folksonomies. There are emerging
approaches for harnessing global tagging to enhance the quality of online
data and to apply some level of control and consistency in their use.
CHAPTER 13
Very large data collections
Overview
This chapter concentrates on aspects of retrieval and management that are particular to
big data. This book originally set out to consider metadata about documents and
document collections, using a wide definition of documents to include images, sound,
museum objects, broadcast material, as well as text-based resources such as books,
journal articles and web pages. Social media activity has been included in this, because
it involves a permanent (usually text-based) record of social interactions or online
behaviour. The type of metadata associated with each of these types of big data will vary
considerably, as will the use to which it is put. Transactional data has largely been
excluded from this scope, unless those transactions relate to documents. This chapter
also describes linked data, an approach that expands the scope of data sets enormously,
because it provides a mechanism for combining data sets from different repositories or
collections – mediated by the internet.
The move towards big data

The move toward big data has been driven by increasing storage and
processing capacity, the establishment of standards for exchange of data and
the requirement of funders to make research data more widely available. This
last factor is based on the idea that publicly funded researchers should make
their data available for further exploitation. It is also driven by regulatory
factors such as those that apply to the pharmaceutical industry. Criticism of
clinical trials data focuses on the selective nature of publication, with the
tendency for some pharmaceutical research companies to publish only data
that favours their products, the phenomenon of ‘missing trials data’

documented by Ben Goldacre (2013) in his book Bad Pharma. The US
government now requires all clinical trials to be registered according to
Section 801 of the Food and Drug Administration Amendment Act, which
came into force in 2017. The registration includes details of documents and
data sets arising from the clinical trial including:
Type
Definition: The type of data set or document being shared.
• Individual Participant Data Set
• Study Protocol
• Statistical Analysis Plan
• Informed Consent Form
• Clinical Study Report
• Analytic Code
• Other (specify)
URL
Definition: The Web address used to request or access the data set or document.
Identifier
Definition: The unique identifier used by a data repository for the data set or
document.
Comments
Definition: Additional information including the name of the data repository or
other location where the data set or document is available. Provide any
additional explanations about the data set or document and instructions for
obtaining access, particularly if a URL is not provided.
(US National Institutes of Health, 2017)
There has been a great deal of commentary about the growth of big data
(Mayer-Schönberger and Cukier, 2013; Davenport, 2014; Kitchin, 2014). It
encompasses many different areas and includes transactional data, research
data collections, unstructured data in an organisational context (mostly
documents), as well as large bibliographic collections. Each of these areas has
its own challenges of complexity, volume and quality. Metadata provides an
important means for accessing information in big data collections, and it
needs to be managed to do so effectively. The metadata status will depend to
a great extent on the nature of the ‘big data’ being interrogated. Metadata is
also needed to manage large data collections. Some aspects such as
preservation, rights management and retrieval are covered in earlier chapters.
VERY LARGE DATA COLLECTIONS 205
What is big data?

The ability to gather and store large volumes of data cheaply has led to an
industry devoted to exploitation of large data sets. Data mining is based on
querying large data sets, sometimes unstructured, sometimes very
heterogeneous, to extract new insights and to develop appropriate business
strategies. It tended to focus on transactional data such as customer purchases
and stock levels in retail outlets. The technology has also been more recently
applied to large text collections to discover patterns that would not otherwise
be evident. However, some of the statistical techniques developed by Shannon
(Shannon and Weaver, 1949) and colleagues are precursors to current text
mining techniques (Bholat et al., 2015).
What is ‘big data’? Davenport (2014, 6–9) talks about ‘big data’ in terms of
volume, variety and velocity. The volume is self-evident; it must be a
substantial collection of data. However, even this aspect can cause problems
because there is no universal agreement about what a large- or high-volume
collection is. The variety refers to the range of data types that may constitute
a ‘big data’ agglomeration. The data may be unstructured – such as text
documents held in a document/records management system – or it may come
from a variety of sources with different data structures. These may be internal
databases or data silos, or may be made up of data from different
organisations. The semantic web is based on the creation of links between
these disparate sources. The third attribute of big data is velocity: is it a
rapidly changing data set – such as social media postings for instance, or
transactional data generated in a shop or an online retailer? If the data under
consideration fulfils these three criteria (and sometimes only two), it can be
considered to be ‘big data’. There is also something about purpose that
characterises some ‘big data’ collections. Very often they are spoken about in
terms of analysis beyond the original purpose of the data. For instance,
documents generated within an office environment may have a number of
different purposes: to document business decisions; to make a record of
transactions for accounting purposes; or to market a product to customers.
These documents may be aggregated into a data set and analysed using
sentiment analysis, or to spot up-coming issues, or to mine for new ideas to
improve profitability. Perhaps just by making big data available it is possible
to generate new applications and potential solutions.
Some of the data (or some parts of the data collection) in a repository may
be structured as a database, in which case identifying the relevant metadata
is more straightforward. The data dictionary or field definitions provide the
metadata structure for the data set. If the data is unstructured, metadata may
have to be created or automatically generated (or extracted) from the data
collection. Metadata can be applied to collections, to individual digital objects
(or documents), and to metadata itself. As more and more organisations are
migrating information repositories to the cloud there are opportunities to
break down silos and offer access to a broad range of data via a common
interface. Text and data mining techniques have been promoted as one of the
benefits of cloud services, alongside resilience, accessibility and cost savings.
However, they also introduce a level of complexity and the need for
description of resources. At the time of writing there do not appear to be any
established metadata standards for describing content held on cloud services,
although services such as Cloud Foundry do suggest metadata for service
development projects on its platform (Cloud Foundry Foundation, 2017).
Commercial services such as Amazon Web Services and Google Cloud
Platform handle metadata about applications and support imported metadata
associated with applications. Daconta (2013) suggests metadata attributes that
should be considered, in very general terms. He suggests that each resource
should have a unique identifier that allows a range of attributes to be
associated with that resource. The attributes (described using metadata) may
be constant (e.g. title, creator, date of creation) or dynamic (such as usage and
other transactional data). Taxonomies may be used to categorise the resources
and linked data provides connections to other resources. However, none of
this is specified in terms of actual metadata standards. Part of the problem is
that cloud services encompass a very diverse range of data collections and
resource types. It is unlikely that a single metadata standard would
adequately address them all. That said, identifiers such as DOI are suited to
wide-ranging resources and linked data has proven to be extremely flexible
and accommodating of different data types. Other researchers have also
recognised the need to manage metadata associated with ‘big data’ (Grunzke
et al., 2014; Sweet and Moulaison, 2013).
The role of linked data in open data repositories

RDF triples are designed for linking data sets on the internet and are used to
illustrate the way in which metadata operates in big data collections. RDF can
be used to express metadata conforming to many different standards. The
use of RDF also opens up the possibility of incorporating data from many
open data sources that are available.
For instance, to build up triples in RDF for a semantic web application, it is
necessary to identify the specific data element and its attributes that can be
used to link different data sets together. So, a book title using an ISBN as the
identifier may link to a bibliographic record with the citation details required
to purchase the book from a publisher’s or bookseller’s database. It might also
link to reviews of the book posted to a social media site, helping an individual
reader or a library to make a purchasing decision.
A key strand of open government initiatives is making data gathered by
public sector organisations freely available to the public. There are two
requirements – the first is to make the public aware of the data sources, by
means of resource discovery sources. The second requirement is to make the
data usable, which can be achieved by providing the data in RDF triples. The
open government initiative in the European Union, for instance, has resulted
in large data collections becoming freely available for use by commercial
organisations, academic researchers and even individuals. This has resulted
in a metadata strategy and recommendations to public authorities responsible
for publishing government data sets (European Commission, 2011). Services
that combine geographic data with public transport data, for instance, provide
live departure and arrival boards for commuters in large cities around the
world. Open government initiatives also improve accountability by making
the operation of government more transparent to their populations.
To make the data sets discoverable, some national governments have
created data portals. For example:
https://data.gov.uk UK
www.dati.gov.it Italy
www.data.gouv.fr France
www.data.gov USA
There are also international portals such as the European Union Open Data
Portal covering over 10,500 data sets, which are described using DCAT
metadata (Publications Office of the European Union, 2017; W3C, 2014a). The
data sets are grouped in the following broad categories:
• employment and working conditions

• science
• social questions
• environment
• economics
• finance
• trade
• production, technology and research.
The European Data Portal, funded by the EU, contains details of over 600,000
data sets from the public sector in Europe (European Commission, 2017a). It
harvests data from other catalogues of data sets, including national catalogues
provided by each of the EU and EEA member governments. The search

interface allows refinement according to facets that include country, topic
category, format of data and licensing arrangements (Figure 13.1). It
advocates use of the Data on the Web Best Practices with suggested metadata
fields for data set providers to include when registering resources on the data
portal (Farias Lóscio, Burle and Calegari, 2017).
Figure 13.1 Screenshot of search results from the European Data Portal © 1995–2016,
European Union
Another service, Data Portals, has details of 520 data portals worldwide and
uses the CKAN software for making data sets available (Open Knowledge
International, 2017). Many of these are local or national portals, some are
subject-specific, and some resources describe ontologies or vocabularies that
can be used for describing data sets. The CKAN system has the following
metadata fields built in:
• title
• unique identifier
• groups
• description
• data preview
• revision history
• licence
• tags – uncontrolled, although they can be organised into tag vocabularies
such as country, composer etc.
• format(s)
• API key
• extra fields.
This is one of a number of data standards used by data portals. Others include
Dublin Core, Project Open Data (POD), Data Catalog Vocabulary (DCAT) and
Schema.org. The US Government Project Open Data analysis of the different
metadata standards provides a useful table for comparison of metadata field
equivalents (Table 13.1).
Table 13.1 Comparison of metadata fields required for data sets in Project Open Data (source:
Federal CIO Council, 2017)
Label POD CKAN DCAT Schema.org
Title title title dct:title schema:name
Description description notes dct:description schema:description
Tags keyword tags dcat:keyword schema:keywords
Last modified n/a dct:modified schema:dateModified
Update
Publisher publisher organisation → title dct:publisher schema:publisher
Contact contactPoint maintainer dcat:contactPoint n/a
Name
Contact mbox maintainer_email foaf:mbox n/a
Email
Unique identifier id dct:identifier n/a
Identifier
Public accessLevel n/a n/a n/a
Access
Level
Data in an organisational context

One of the major challenges facing large organisations is the diversity of data
sets generated and the inevitable rise of silos of information. This may be for
historical reasons – particularly if the organisation is the result of a merger or
take-over. There will be legacy databases that need to be reconciled. There
may be very good operational reasons for keeping separate data sets – for
instance, protecting sensitive personal data from unauthorised access. More
often the data sets arise from specialist software applications used to run
operations such as finance, customer relationships, marketing, logistics,
human resources, transactional processing – to name a few. Each of those
specialist systems will have its own data structures and will handle data in
ways that are specific to that application. Sometimes there will be common
data standards for interchange of data between different systems, but the
internal handling of the data may use proprietary standards. For instance, in
the library management field, MARC 21 is often used for exchange of
catalogue data between systems, but the internal data format may be more
complex to reflect internal management and auditing requirements.
The bigger challenge is unstructured content. The availability of powerful
software and services such as Google Enterprise Search, SharePoint and
Oracle allows organisations to search very large repositories of documents
that may have minimal structure. Although retrieval may be by means of
probabilistic searching and ranking rather than matching exactly to Boolean
operators, there are many challenges including duplication of content (all too
common and exacerbated by the lack of reliable retrieval), different
terminology or even language to describe the same concept or topic and
mixed media content. Individual documents may have some structure,
imposed by the software system or the organisational style manual, but there
may be little direct correlation between different applications. Even within
the same family such as Microsoft Office, each application has its own
document properties interface. The challenge is then to search across all these
different formats.
Several researchers have considered the use of metadata as a solution to
retrieval and management of unstructured document and content collections
within an organisation. Initiatives include the DMS Mark-up Language, a
metadata standard for multimedia content management systems (CMS), and
the Darwin Information Typing Architecture (Paganelli, Pettenati and Giuli,
2006; Anderson and Eberlein, 2015; Sheriff et al., 2011; Bailie and Urbina,
2012). The success of these approaches depends on the degree to which they
can be incorporated into document or enterprise information management
systems. Sheriff et al. (2011) identify the following sources for CMS metadata,
which could form the basis for a general approach to organisational content:
• generic metadata standards

• content-dependent metadata
• industry-dependent metadata
• custom metadata (organisation-specific).
Social media, web transactions and online behavioural

advertising
Metadata for social media brings up a number of issues such as personal data
and privacy, as well as online behavioural advertising, which depends on
metadata. Metadata associated with communications and messages are
another key area which has an impact on privacy and civil liberties. The past
is littered with cases of abuse of personal data by the state (for ethnic cleansing
and genocides, for instance).
This book started with a description of the Snowden revelations about the
wholesale collection of metadata associated with telephone and e-mail
communications (Greenwald, 2014). By his account, the metadata was actually
more revealing than the content of messages, because of their standardised
form, which would facilitate analysis of large volumes of data. Metadata
collected by the US National Security Agency (NSA) included details of
telephone numbers called, duration of calls, names of line subscribers along
with location data.
Social media services account for the generation of huge volumes of
content, as text, pictures, moving images, instant messaging, sound, etc. The
files attached to social media postings such as images and documents usually
have embedded metadata that is relevant to that file format. Standards for
different media include MPEG, JPEG, and Dublin Core, discussed earlier in
Chapter 4. Social media have also adopted standards for metadata to facilitate
handling of postings. For instance, the Open Graph Protocol (described in
Chapter 4), developed by Facebook is widely used to describe the relationship
between postings and between individual users of social media. Standards
such as VIAF may also be useful for describing content.
A closer look at social network providers reveals a more detailed approach
to metadata, designed to facilitate use of data about users and their online
behaviour by applications that call out data from the social networks (APIs)
and by online advertisers.
Although Facebook has recently placed some restrictions on the metadata
that is available to third parties, it is still possible to glean a great deal of data.
For instance, Facebook gathers to up to 98 data elements associated with
individual users to build up profiles for digital advertising (Dewey, 2016).
The digital advertising business drives a great deal of activity on the internet.
Social media companies (such as Facebook and LinkedIn), search engine
providers (such as Google and Yahoo!) and large commercial and government
websites gather data about web activity and use this for marketing, promotion
or public information. Aggregators bundle up usage data and sell it to
advertisers or agencies so that ads can be directed to targeted audiences.
Figure 13.2 represents a simplified overview of the different agents involved

in digital advertising.
Ad
Exchange
Advertiser/
Agency Publisher
Ad
Network
Figure 13.2 Agents involved in delivering online ads to users
A publisher of a website may belong to an ad network with a number of other

publishers. Through the ad network they serve targeted ads to advertisers (or
their agents) in return for a fee. The ad exchange provides a mechanism for
trading ads in bulk so that advertisers get the best price and publishers are
able to sell available ad space on their websites. They gather information on
target audiences by placing a cookie on the browser of customers or visitors
to their website. This tracks subsequent activity and allows the advertiser (or
advertising agency) to build up a profile of that particular user’s interests.
This then allows targeted advertising by serving ads at the appropriate point
in a browsing session.
Research data collections

Research data collections, such as the UK Data Archive (University of Essex,
2017) and the National Climatic Data Center (National Oceanic and
Atmospheric Administration, 2017), have been around for some time. With
the move towards open data, many funders have started to require research
data to be put in the public domain so that it is available to other researchers
and the wider research community. Some of these are subject-based, some
are incorporated into institutional repositories, and some are listed as part of
open government initiatives. Metadata is used to make these data sets
retrievable and to some extent to manage them.
The Open Discovery Initiative has recommended a minimum core set of
metadata in repositories exposed to index-based discovery systems. There is
a strong emphasis on bibliographic and media materials (Table 13.2).
For data sets that are dynamic, there is the challenge of keeping track with
Table 13.2 Core metadata elements to be provided by content providers (from Open
Discovery Initiative Working Group, 2014, 16)
Field Name Definitions

Title The main title of the item
The author(s) of the item. Individual authors should be listed in
Authors
lastname, firstname order
Publisher Name The name of the publisher of the item
Volume Volume number of the resource, where applicable
Issue Issue number of the resource, where applicable
Page(s) Page numbers of the resource, where applicable
The date of publication. For a serial run, coverage dates included
Date/Date Range
for the serial
One or more standard identifiers for the print or online version of
Item Identifier the item (e.g. ISSN, OCLC number, ISBN, DOI, etc.). The identifier
should be preceded by a label indicating the type of identifier
Describes the publication or serial of which the individual item is a
Component Of Title part (e.g. for journal articles, the serial title; for tracks on a CD, the
album title; etc.)
Provides a standard identifier for the component title defined
Component Of Title
above (e.g. ISSN, OCLC number, ISBN, DOI, etc.). The identifier
Identifier
should be preceded by a label indicating the type of identifier
Item URL Either an OpenURL or a direct link for the specific item’s full text
To comply with the NISO Open Access Metadata and Indicators
Open Access (OAMI) group’s recommendations, if an item is open access, this
Designation status should be indicated with ‘free_to_read’ and otherwise left
blank. See www.niso.org/workrooms/oami
A yes/no statement describing whether the content provider

makes this item available in full text (or for non-print media, a full-
length or high-resolution version) to the DSP for the purpose of
Full Text Flag
indexing. It is expected that this will be disclosed by DSPs to
libraries in future when describing indexing coverage for a title or
collection
Intended to be used to identify whether the content being

described is textual, a visual recording, a sound recording, etc. The
Content Type* textual descriptors from the controlled list established in the
MARC 21 Type of Record position (06) of the Leader field is
recommended to be used for this field’s content
Intended to be used to indicate whether the nature of the content

being described is monographic, serial, a component part,
Content Format collection, etc. The textual descriptors from the controlled list
established in the MARC 21 Bibliographic Level position (07) of the
Leader field is recommended to be used for this field’s content
* It is recognized that many content providers merge Content Type and Content Format
in their systems. Providing separate fields for this data is preferred, but the current
practice of a single field may continue if separating the data is too burdensome.
those changes – a problem of lack of fixity (to use records management

terminology). There are also problems of consistency. One study of the Dryad
data repository (which contains biology and ecology data sets) highlighted the
problem of lack of consistency of use of data elements (Rousidis et al., 2014).
For instance, the way in which names are expressed in the dc.creator data
element is inconsistent and error-prone. The paper suggests that an author
identity system such as ORCID would overcome this problem. They even
suggested that a link to the ORCID record would allow for updated author
details to be automatically propagated to the metadata record. This may not be
such a good idea, because the details at the point of creation of the record may
be more relevant than the updated details. Providing a link to the updated
record, however, does allow for forward tracking. The same authors also
highlighted problems with variable date formats (if not system-generated) and
suggest a validation algorithm. They also looked at dc.type and suggested that
a controlled drop-down list should be used to guide the data entry personnel.
There are three metadata management approaches highlighted in Chapter 11.
De Waard (2016) introduces the idea of a research data needs pyramid,
which nicely illustrates the different aspects of research data management
and processes that need to be considered. Figure 13.3 groups these require-
ments into three types: Saved, Shared and Trusted.
The model developed by de Waard refers to the FAIR principles developed
SUCCESSFUL
DATA
Reusable
C
Trusted
Reproducible
Reviewed
Comprehensible
Citable
Discoverable
Shared
Accessible
Preserved
Stored
Saved
Figure 13.3 A ‘pyramid’ of requirements for reusable data (de Waard, 2016)
by the European Union, which require that research data is Findable,

Accessible, Interoperable and Reusable (Wilkinson et al., 2016). The focus of
this is primarily on metadata quality and standards and this is a good
demonstration of the key role that metadata plays in the management and
use of research data sets. Findable means that the resource can be discovered
via user queries. Metadata may be quite structured to allow retrieval via
different criteria such as creator, dates, geographical location, or it may be
descriptive of the topic or subject, which could be via a controlled vocabulary.
Accessible means that the format is known to the user and that they have
appropriate access rights. It may also relate to ability to perceive (such as
special interfaces for people with visual impairments). Interoperability
ensures that information about the data (metadata) can be utilised on different
systems or may be harvested via OAI-PMH protocols, for instance. It also
allows for the data itself to be used by different applications or APIs. Linked
data expressed in RDA is a good example of this type of interoperability (RDA
Steering Committee et al., 2016). The metadata should be reusable, preferably
conforming to a standard and with appropriate licensing arrangements to
allow re-use. One class of research collection is the compilation of web-scale
discovery systems (such as Ebsco Discovery Service, Primo from Ex Libris,
Summon from ProQuest and WorldCat Local from OCLC), which harvest
metadata from a number of sources. This raises metadata quality and
interoperability issues, including: differences in granularity; different
metadata standards; ambiguous identities; differences between relevancy
ranking algorithms; duplication of resources; different metadata harvesting
schedules; inclusion of open access material; and differing licensing
agreements (Breeding, Kroeger and Sandy, 2016).
DataCite
The DataCite metadata standard is for describing and disseminating research
data and has been developed with strong input from the research and
academic communities. Its goals are to facilitate access to research data via
the internet; make research citable in the scholarly content; and to support
data archiving (DataCite Metadata Working Group, 2016). The data sets
described by DataCite metadata include numerical and other types of
research data. DataCite is a member of the International DOI Foundation,
which means that its member institutions mint DOIs for their data clients.
There is a small mandatory set of metadata required to register research data:
• DOI
• Title
• Creator
• Publisher
• Publication Year
• ResourceType
Recommended properties:
• Subject
• Contributor
• Date
• RelatedIdentifier
• Description
• GeoLocation
Optional properties:
• Language
• AlternativeIdentifier
• Size
• Format
• Version
• Rights
• FundingReference
One of the challenges of maintaining dynamic data sets is the need to have a
stable referencing system, but to be able to incorporate additional data as it
is produced. Researchers can use the RelatedIdentifier and Version data
elements to specify updates to a data set.
Institutional repositories
Institutional repositories grew rapidly in the academic sector in the early
2000s. Academic institutions around the world realised that there were
benefits from putting their research outputs and some research data and
primary research resources on a system that would facilitate use and sharing
internally and by the global academic community. They have become an
important part of the preparation of bids for funding and for assessment of
research quality. Access to research repository data is available via directory
services such as OpenDOAR (University of Nottingham, 2014). This type of
service allows users to search across institutional repositories around the
globe and to filter results by criteria such as subject, institution, software used
and country. It is possible to search by institution, by collection or by
individual item. Access to some items may be restricted because of copyright

or licensing requirements of the intellectual property holders. So, for example,
some journal publishers will allow pre-prints of articles or even post-prints of
articles to be available to members of the institution that the research originated
from, but not to external users of the repository. Although there are many
different software platforms for institutional repositories, a few dominate the
market. In the early days many institutions developed their own systems. Some
commercial providers also developed institutional repository software with
various degrees of integration with their library management systems. This
allows the potential of an integrated search interface for members of an
institution. The most common software applications used for open repositories
(of the 3320 repositories listed on OpenDOAR) are: DSpace (1473), EPrints (452),
Digital Commons (159), and Opus (80). The distribution of repositories is
strongly skewed to Europe (1503), Asia (671) and North America (602)
(University of Nottingham, 2014). One of the features of institutional
repositories is their ability to interoperate. This means that there has to be a
common exchange format, even if repository software applications have their
own internal standards or the institution catalogues material to other standards.
Dublin Core is the commonly accepted minimum requirement for exchange of
metadata via OAI-PMH. Other, more detailed metadata standards such as
MODS and MARC 21 are also widely used. Repository directory services such
as OpenDOAR are metadata harvesting systems. They regularly interrogate
the metadata stores of the institutional repositories that they know about and
have access to and update their own central metadata store. This allows users
a single interface to all the indexed institutional repositories and provides rapid
retrieval. Pointers in the metadata store then allow access to individual records.
Index-based search services (metasearch) are taking over from an earlier
generation of federated search services such as Copac in the UK. The UK’s
National Bibliographic Knowledgebase will replace Copac and may adopt an
index-based approach such as that used by WorldCat, the global service
operated by OCLC. These architectures are illustrated in Figures 13.5 and 13.6
(pages 218 and 219). Both of these architectures offer advantages over silo-
based searching (Figure 13.4 on the next page), where a user interrogates each
individual repository in turn. The federated search services may de-duplicate
search results to provide a consolidated list with pointers to the original
sources. The index-based search service offers better performance and rapid
retrieval, because the user does not have to wait for responses from many
individual repositories to obtain comprehensive results.
The OAI-PMH (Open Archives Initiative – Protocol for Metadata Harvesting)
standard is still widely used for metadata harvesting, although it has been
superseded by ResourceSync (Lagoze et al., 2002; Open Archives Initiative,
Source
Repository
Source
Repository
User
Source
Repository
Source
Repository
Figure 13.4 Silo-based searching
Source
Repository
Source
Repository
Federated
User Search
Service
Source
Repository
Source
Repository
Figure 13.5 Federated search service
2017). The OAI-PMH is optimised for harvesting metadata from catalogues of

cultural and bibliographic objects. The ResourceSync is intended to be a more
general harvesting protocol that allows for harvesting of metadata about digital
objects, as well as the digital objects themselves. It can exploit sitemaps using
metadata about the URL such as the location timestamps of last modification.
An initial baseline synchronisation is necessary, but thereafter the destination
system polls sources for metadata about changes that have taken place over a
given time period and then compares with previous updates to determine
Source
Repository
Source
Repository
User Metadata
Index harvesting
Source
Repository
Source
Repository
Figure 13.6 Index-based discovery system
whether it is necessary to download updated material. The system is

unidirectional in that the destination service pulls the data from the source,
but the source does not receive data from the destination service.
Conclusion
Recent growth in interest about ‘big data’, data mining and very large data
repositories has been paralleled by developments in metadata harvesting and
search architectures. Metadata standards such as D-CAT have been developed
to help structure information about large collections (Open Archives
Initiative, 2017). This enhances interoperability between systems and helps
to address some of the issues of data quality and consistency. Even in areas
where there is formal control of the metadata, such as bibliographic collec-
tions, the variety and variance in the use of cataloguing standards creates
challenges for services that span different collections or institutions. This
problem is magnified when searching across data repositories where
individual data collections may be highly structured but there is little or no
commonality between collections. Within organisations there is a separate
problem of bringing together data and information items from multiple silos
in different formats and often with little consistent internal structure. A good
example is the proliferation of word-processed documents that may be made
available via a document or records management system. These tend to rely
on categorisation by purpose or tagging by authors. Powerful text retrieval
techniques have improved access to documents, but do not fully address the
problems of consistency, precision or recall. Social media generates huge
volumes of transactional data and this is used to sell advertising to users. The
digital advertising industry depends on metadata associated with user online
transactions to build up these profiles as well as the personal data that is
gathered or inferred from users’ social media profiles. Research data
collections are a major enterprise, although this tends to be publicly funded.
Standards such as DataCite produce metadata that allows data sets in
repositories to be discoverable. Institutional repositories tend to focus more
on bibliographic data associated with research publications, although some
also contain primary data gathered in the course of research.
CHAPTER 14
Politics and ethics of metadata
Overview
This final chapter considers metadata in a political context. It considers three aspects of
its role in society and speculates on possible future developments. There is the ethical
strand, an increasingly important considerations for those involved anywhere in the
information communication chain (Robinson, 2009). It also considers where the power
lies in both professional and subject domain terms and which professional groups are
best equipped to develop and implement metadata standards. This section speculates
on the role of metadata in the creation of new knowledge – a holy grail that has so far
eluded the most advanced machine learning environments. Finally it considers the
practicalities of funding. There is a huge industry dependent on metadata about online
transactions, for instance. This forms the basis of digital marketing and the revenue
streams for some of the largest incorporated companies, such as Alphabet (Google) and
Facebook. This also raises the issue of who pays for the creation of new metadata
standards, and who funds the creation of metadata on a massive scale (all those
digitisation projects). Throughout, the chapter speculates on the future development and
role of metadata.
Ethics
An examination of the role of metadata raises many issues about privacy,
security, ownership and control. It also raises issues about the digital divide
and its possible role in making information accessible to wider audiences. It
has the potential to empower the marginalised, hold government to account
and improve individuals’ quality of life. Understanding metadata is
important in information literacy and helps individuals to navigate the

turbulent sea of opinions, fact-free news and propaganda.
Privacy and ownership

In the context of social media, online usage and communications, metadata
has become personal. Much of the data about internet activities and
transactions is about personal activity and online behaviour. According to
our original definition metadata describes an information object, whether that
be raw data or more descriptive information about an individual. This is
important because the treatment of metadata has become a political issue.
Personal data, especially data that reveals opinions, attitudes and beliefs, is
potentially very sensitive. Use of this personal data by service providers or
by third parties can expose users to risks such as nuisance from unwanted
ads, harassment from internet trolls or fraud through identity theft, if the data
is not held or transmitted securely. Countering this, many digital advertisers
would say that because the data is aggregated it is not possible to identify
individuals – i.e. the data is anonymised. However, even the official guide-
lines by bodies such as the ICO suggest that anonymisation of aggregated
personal data is not foolproof (Information Commissioner’s Office, 2012;
Chen and Li, 2013; Narayanan and Shmatikov, 2009). It may be reversible, or
personal identity may be revealed when combined with easily accessible data
such as that from an electoral register.
Some commentators have suggested that a way around the problem of use
of personal data is to strengthen the control an individual has over his or her
personal data (Bond, 2010). The idea of ownership of data extends to those
who have access to it and even remuneration for being allowed to use that
personal data. This is an idea that has taken hold since the proposal was put
forward that personal data should be treated as a new asset class (World
Economic Forum, 2011). It could be the basis for a new, more balanced
relationship between individual users and service providers. However,
without legislation and the co-operation of the main digital content providers,
it is difficult to see how such an approach would take root.
Security
Metadata has become a political issue. Anyone who had asked the question
‘What does metadata matter?’ prior to 2013 will have been startled by the
revelations about the US National Security Agency’s routine downloading of
metadata about telephone conversations that involve non-US citizens
(Greenwald, 2013). The Fourth Amendment to the US Constitution protects
POLITICS AND ETHICS OF METADATA 223
‘The right of the people to be secure in their persons, houses, papers, and effects,
against unreasonable searches and seizures’ (United States, 1791). A lot hangs
on the interpretation of privacy, as Solove (2011) has so eloquently discussed
in his book Nothing to Hide. Individuals are monitored continually by CCTV
cameras, through their communications and particularly their online activity,
by security agencies, crime prevention and investigation agencies and by the
digital advertising industry. They all exploit metadata that reveals information
about us as individuals, whether it be online grocery shopping or hidden
activity on the ‘dark net’. The metadata about these activities and transactions
are the surrogate resource that is used to filter and aggregate data to find and
act on leads. Privacy International has identified the following types of
metadata that is gathered or could be gathered by security agencies:
Under the traditional definition of metadata, this information about our

communications [surveillance metadata] would include:
• the location that it originated from, e.g. home address of the telephone,
subscription information, nearest cell tower
• the device that sent or made the communication, e.g. telephone identifier,
IMEI of the mobile phone, relatively unique data from the computer that sent
a message
• the times at which the message was made and sent
• the recipient of the communication and their location and device, and time
received
• information related to the sender and recipients of a communication, e.g.
email address, address book entry information, email providers, ISPs and IP
address, and
• the length of a continuous interaction or the size of a message, e.g. how long
was a phone call? how many bits in a message?
(Privacy International, 2017)
Addressing information inequality

Another aspect of ethics is the growing information inequality. The digital
divide means that there are divisions between nations as well as within
societies (Norris, 2001). There is a large body of digital citizens leaving behind
those excluded from society by poverty, age, lack of education. Metadata has
a role in making information accessible to groups excluded, for instance, by
physical disability. Schemes such as the W3C Web Content Accessibility
Guidelines (WCAG) provide a framework for describing different aspects of
accessibility for websites (Caldwell et al., 2008).
Metadata also has a key role in information discovery: getting information

to those that need to know. Chapter 6 looked at information retrieval and the
techniques used to make information accessible (such as metadata generated
from controlled vocabularies). Information literacy is one way of addressing
this and it could be argued that metadata awareness should be a part of
information literacy.
Improving discovery
Effective presentation of metadata enhances its usability. A key aspect of
presentation is the ease of navigation and searching. The navigation system
and search facilities have to accommodate the needs of different kinds of
users. Some people interact with systems when they create a new document
and they will need to create the metadata. Other users will be primarily
interested in using metadata to retrieve electronic resources as searchers.
When entering metadata an author may need access to a controlled
vocabulary, in order to select appropriate keywords. The terms can be
presented in a number of ways:
• A drop-down list – This method is suitable for short lists of terms, or

where there are a very limited number of possible options. It allows a
user to immediately see the full range of choices available, and so help
them to select the most relevant items.
• A navigable classification scheme – This is more appropriate for a longer list
of controlled terms, where they can be categorised according to a
classification scheme. Some applications that support thesaurus
relationships can display terms in a hierarchy. This allows users to
navigate through the classification until they find the appropriate
term(s).
• A search – The user enters the term into a search box and is presented
with an alphabetical listing centred on the point of the search term
entered. They can then browse the area and select the appropriate terms.
• An automatic indexing system – Many systems offer automatic indexing
capabilities. In effect the system examines the resource being described
and suggests appropriate indexing terms. This technique can be used for
keywords or subject descriptions, where the system refers automatically
to a thesaurus of controlled terms as described earlier.
General users or searchers can interact with electronic systems using the first
three options to identify relevant search terms or selection criteria for the
metadata, and ultimately the resource that is described by the metadata.
User education
Although there is increasing awareness of the existence of metadata, many
users do not understand how metadata works. At the most basic level there
is a need to identify which metadata fields are available for searching. If the
content of the fields are controlled (either by cataloguing rules or by use of
controlled vocabulary), the user needs to know where they can browse the
available keywords or terms. More sophisticated searching will require an
understanding of ways in which search queries can be combined so that, for
instance, it is possible to search for the author and title of a book on Amazon,
or to search for the subject category and date of creation of a web page on a
government portal. Commentators such as Phil Bradley have written
extensively about this (Bradley, 2013).
The idea of the selective content has been around for some time, but was
brought to prominence in The Filter Bubble (Pariser, 2011). Unexpected
electoral results in 2016 such as the Brexit referendum in the UK and the
Presidential elections in the USA highlighted the problem of pollsters
themselves living in a filter bubble. This made it difficult for them to
understand or even acknowledge dissonant views from parts of the electorate
that they do not usually consider significant. The other phenomenon has been
‘fact-free news’ which has affected the political discourse in the USA, Russia
and the European Union in particular. In a former age this might have been
called propaganda. Material is easily generated in response to the economic
model that depends on driving traffic to websites and thereby generating
income from advertising revenue. The wilder or more outrageous the
assertion the higher the traffic. This turns out to be an effective political
campaigning strategy as well. The distinction between speculation and
established fact backed by evidence is still as stark as it ever was, but the usage
of these two types of approach is becoming blurred. Services such as
Facebook do not discriminate between evidenced material and
unsubstantiated speculations when they serve up the toxic mix of fact and
fictions to their subscribers.
Service providers tailor content to suit our opinions as revealed by previous
online activity. Ad blocking and cookie blocking limits the amount of data
gathered about your online activity. Blocking also stops some of the benefits
of cookies, such as continuity of sessions, tailoring of content and the ability
to make purchases online.
What role does metadata play in all of this? At least part of the basis for
selection of news stories to serve to users is based on matching a profile of
past interests and internet behaviour with new content. It seems that we are
more comfortable with news and commentary that reflects our own opinions
and preconceptions (Norris, 2001, 18–19). This is why so few liberals read
conservative newspapers and vice versa. Whether the metadata is

automatically generated or is manually applied to a news story (perhaps both
occur), the net effect is that we are served content that reinforces our opinions
and prejudices. Selection of news stories may be based on metadata associated
with them – such as the source, the author, and key words and tags associated
with the story. On platforms such as Twitter, user-allocated hashtags also play
an important role. Hashtags arise spontaneously when someone notices a
trend or decides to create one. There is no control, and a hashtag may have
multiple meanings depending on who is using it.
For example, the terrorist attack on the UK Parliament in Westminster on
22 March 2017 prompted a lot of Twitter activity using the #Parliament
hashtag. For that time and context the hashtag was sufficiently specific, for a
search two days later to yield tweets mainly about that incident. However,
on another occasion the same hashtag may have an entirely different
meaning.
The following metadata is available for scraping from the Twitter stream:
Twitter handle (of tweeter)

Twitter handle (of correspondent)
Relationship
Relationship date
Tweet
URLs in tweet
Domains in tweet
Hashtags in tweet
Tweet date
Twitter page for tweet
Imported ID
In Reply To Tweet ID.
Fields were extracted from Twitter using NodeXL social analysis software
(Hansen et al., 2011).
Power
Who owns the metadata space?
There have been many successful collaborations, such as the one between
librarians and the IT community to create Dublin Core as a metadata standard
that could be applied to web resources. However, there are distinct
communities with their own perspectives on metadata, such as: librarians,
data scientists, database designers, geographic specialists, geneticists and

statisticians. Chapter 1 considered the history of the concept of metadata and
its emergence from early attempts to catalogue collections of scrolls, such as
the catalogue of the Great Library of Alexandria more than 2500 years ago
(Gartner, 2016, 16–18). We consider not only the professional groups with
specific concentrations of skills and experience, but also different subject
domains that have been particularly preoccupied with the issue of describing
information or data. We ask the question: ‘Who is best placed to develop,
implement and use metadata in the future?’ Librarians, information
architects, database developers, and knowledge managers have all staked a
claim on metadata. Big data specialists such as data scientists, business
analysts and data modellers also have an interest in metadata. Each of these
groups brings different skills to the metadata field and has its own perspective
on how best to utilise and manage metadata. The development of metadata
by the geographical and spatial science communities arose from their
intensive use of large volumes of data in a variety of formats and for a variety
of purposes. This meant improving the precision of data attribute
descriptions. Geographic data is very diverse and may be used to delineate
an area of land, describe an orbit around a planet or pinpoint a signal. The
metadata allows systems to handle and interpret geographical or spatial data
in an appropriate fashion. The precision of location data may vary from a
country, from a street address or to very precise co-ordinates in a defined
area. Location information pervades many other areas of activity, so that, for
instance, the majority of local authority data has a geographical component.
Metadata standards such as ISO 19115 (ISO, 2014a) and INSPIRE (European
Commission, 2017b) arise from the geospatial community.
Data dictionaries associated with database design and management are
another form of metadata. Although it is beyond the scope of this book, which
concentrates on data about documentary resources, some of the data
modelling techniques that they use are relevant (Hay, 2006).
Subject-based communities that make intensive use of large volumes of
data have developed metadata standards. These include genetics researchers,
statisticians and clinical researchers.
Professionals of the future

In the previous edition of this book I speculated about the growth of a
metadata community. Although activity in this area has expanded, a single,
cohesive community has not emerged. If anything, it has fragmented further.
There is a strong emphasis on bibliographic information and there have been
some significant developments in cataloguing and resource description with
the adoption of RDA (Resource Description and Access). There have been
separate initiatives on identifiers – particularly those to do with people, and
separate development of metadata used in the social media space. Although
parts of the archives and records management communities have moved
closer together, they continue to remain independent of the LIS community
at large. It could be argued that the role of archives and records collections
are very different and that therefore the descriptive requirements and
management needs are different. Even within the bibliographic community,
the book trade has developed its own standards for metadata to facilitate e-
commerce. At this juncture it seems that this fragmentation of metadata will
continue and that a coherent metadata community is unlikely to emerge.
Looking to metadata initiatives in other areas, such as the geospatial
community, we see even further fragmentation and the possibility of a
common framework for metadata receding over the horizon as the flotilla of
initiatives disperses.
The more dynamic parts of the information disciplines are changing to
incorporate metadata as a part of their range of knowledge and skills.
Metadata is an established part of many i-school syllabuses and other LIS
academic courses. Job ads often feature the word ‘metadata’, as in ‘metadata
librarian’ where previously ‘cataloguer’ might have sufficed. This suggests
wider awareness of metadata and an appreciation of its role in LIS. The
content management community has started to consider a systematic
approach to documentation and use of metadata standards such as DITA
(Bailie and Urbina, 2012). Although it is beyond the scope of this book, it is
an interesting area to watch and it may have an impact on the adoption of
metadata standards generally.
Exploitation
The major search engines have been another focus of metadata activity as new
products based on the semantic web emerge. The Google Knowledge Base
and the development of Schema.org are examples. Open data initiatives
around the world have resulted in the development of new products and
services that combine data sets. Linked data technology (for example using
RDA triples to describe data elements) has facilitated the combination and
exchange of data. However, consistency of descriptions is a major problem.
If different encoding schemes are used for the content of data elements, there
may be a problem linking them or with information retrieval later on.
Nonetheless, there has been rapid growth of this sector and that seems set to
continue. The first linked open data set was published in 2007. By the end of
that year there were 28 data sets available which had grown to 203 data sets
by 2010 and 570 by 2014. In February 2017 there were 1139 linked open data
sets available (Abele et al., 2017). Knowledge creation has been an ambition
at least since the 1980s, when the Alvey Programme in the UK and Japan’s
Fifth Generation Computer Systems project were in full swing (Rutherford
Appleton Laboratory, 2015; ICOT, 1999). The development of the semantic
web in more recent years has provided an avenue for linking individual data
elements based on their meaning. The EU-funded LOD2 project has explored
advances in semantic web developments but has not addressed the creation
of new knowledge (Auer, Bryl and Tramp, 2014). It may be that knowledge
creation cannot be separated from consciousness. If that is the case, then it
would seem that the early dream of systems that could exploit existing data
sets to create new knowledge is still a long way off. Whatever the agent
(human or artificial intelligence), the idea that metadata could be used to
navigate existing knowledge as a first step to synthesising new knowledge
and insights remains an attractive one.
Money
Who pays?
One of the major challenges for any public initiative is finding the funding
and resources to make it happen. Many public bodies are good at creating
initiatives that address an issue of the moment, but then attention moves on
to other topics. The infosphere is littered with abandoned databases,
protocols, frameworks and information services. Creating metadata is
expensive and research continues into the automatic generation of metadata.
Approaches include automatic indexing, metadata extraction from data sets
and use of specific tools to generate metadata automatically (Golub et al.,
2016; Greenberg, 2009; Park and Brenza, 2015).
Volunteer cataloguing is an alternative approach and there have been
useful initiatives where a few dedicated individuals have done an enormous
amount of indexing and cataloguing. Some clever initiatives have also
allowed general users to contribute to indexing materials. This is particularly
interesting for images and to some kinds of manuscript which cannot easily
be converted into coded text with current technologies (Konkova et al., 2014).
For instance, the British Library Labs enables volunteers to transcribe images
of catalogue cards into machine-readable records by crowdsourcing tasks
(British Library Labs, 2017).
Standards
The standards development process is often prompted by the proliferation of
incompatible systems that arise in response to a practical problem. As
accepted good practices emerge, standards are often developed to codify that
practice. Standards then provide a common approach. This is fundamental
to the successful operation of metadata systems.
One of the most widely used metadata schemes, the Dublin Core Metadata
Element Set is now an international standard, ISO 15836 (ISO, 2009b). It
provides a starting point for many application profiles developed by specific
communities and individual organisations. It forms the basis of many other
standards, such as FOAF, AGLS, eGMS, Europeana Data Model and Pundit
(National Archives of Australia, 2008; e-Government Unit, 2006; Brickley and
Miller, 2014; Net7 srl, 2017; Europeana Labs, 2016).
Standards development is a helpful way of negotiating a system that all
parties can work with. This is evident in the publishing industry and book
trade, which is made up of conflicting and competing interests and yet has
co-operated to develop the ONIX metadata system, because of the need of
the different parties (book retailers and publishers) to exchange data
(EDItEUR, 2014). Another example is the evolution of the national MARC
standards into MARC 21, a unified standard adopted internationally (Library
of Congress, 2017c). Standards development is often seen as a common good.
Publicly funded organisations such as the Library of Congress in the USA or
Jisc in the UK often lead standards development and provide the support and
infrastructure for implementation. Commercial organisations such as data
service providers and systems providers get involved to influence eventual
standards that are used to define the market. A second source of standards
development comes from commercial organisations that want to develop
proprietary standards which tie customers into their products or services.
Some of these, such as Adobe’s PDF format, have reached a wider
community. One of the tasks of bodies that sponsor national and international
standards is to reconcile these different interests to produce a standard that
can be widely adopted. This is an expensive process, which may be funded
by private sector organisations with an interest in the standard. Professional
bodies, trade associations and public bodies (particularly regulators) will all
have a say in standards development and ratification. In this book we have
seen a wide range of organisations involved in standards development,
including those listed in Table 14.1 opposite.
Re-examining the purposes of metadata

The first edition of this book speculated about the purposes of metadata,
Table 14.1 Metadata standards development
Consultative Committee for Space Open Archival Information System

Data Systems (CCSDS)
DCMI Dublin Core
EDItEUR Metadata for the book trade
Facebook Open Graph Protocol
IEEE Computer Society Metadata for learning objects
International Council on Archives Archive description
International Federation of Library Bibliographic description
Associations (IFLA)
International Organization for Metadata and records management
Standardization standards and encoding schemes
International Telecommunication Object identifiers
Union
Library of Congress Bibliographic standards
National Information Standards Open Archives Initiative
Organization (USA)
OWL Working Group Ontology language
PREMIS Editorial Committee Preservation metadata
RDA Steering Committee Bibliographic description
W3C Internet standards
which continues to be an overarching theme of this edition. The original

model identified five main purposes of metadata. This has evolved into a six-
purpose model that includes information governance as a distinct purpose:
1 resource identification and description

2 retrieving information
3 managing information resources
4 managing information rights
5 supporting learning, research and commerce
6 information governance.
The original five-point model described metadata in terms of its purposes.

Although individual data elements may be used for more than one purpose,
the purposes themselves remain distinct. So, for instance, there is a clear
distinction between the title of a book, as a description of that book, and using
words in the title for information retrieval. The model provides a way of
examining metadata across a wide range of different application areas. A
closer look at the new, six-point model demonstrates why it works and
provides some indication of how it might develop in the future.
Purpose 1: Resource identification and description (Chapter 5)

Identification is seen as one aspect of description and is fundamental to other
purposes. An identifier provides a ‘handle’ for a digital object and other
information containers such as books, so that other processes such as retrieval,
electronic trading or rights management can be performed.
The range of things being described using metadata will be extended. At
present metadata is applied to works ranging from books to music and works
of art. It is also applied to data collections and text-based electronic docu-
ments. There is some interest in developing metadata schemas to describe
knowledge, especially tacit knowledge (the knowledge in people’s heads).
This suggests a view that people can be regarded as information resources or
repositories. At the time of writing no major metadata standards had been
developed for knowledge management, although some proprietary systems
may have data dictionary definitions that could form the basis for future
metadata schemes for knowledge. As discussed earlier, a key purpose of
metadata is the ability to identify what is being described. Standards for
identifiers can be incorporated into metadata as a way of controlling the
content, and to facilitate linked data applications. Examples include ISBNs
for books, Digital Object Identifiers (DOIs) for text, and International
Standard Name Identifiers (ISNIs) for people.
In the library sector, there are competing or overlapping identifier systems
and it would be reasonable to anticipate some consolidation in this area, with
the emergence of a single, universal identifier system for all kinds of entity.
The problem is determining the level at which the identifier should be
applied. For instance, a work can have its own identifier, so can individual
editions of a book, each with its own ISBN, and so can individual items in a
collection – typically identified by an accession number, a bar code or an RFID
(radio frequency identifier) chip.
Purpose 2: Retrieving information (Chapter 6)

The second purpose of metadata is one that has attracted more commentary
and speculation than any other. It is the focus of a great deal of the work on
metadata standards for web resources, such as Dublin Core. The retrieval
purpose ties in the description as a means of evaluating a resource when it
has been retrieved.
Image retrieval continues to present challenges of interpretation. For

example: What is the most significant element of a picture? What does it
mean? Automated image indexing depends on shape, colour and texture, but
does not address the ‘meaning’ or interpretation of an image, although some
advances in image recognition technology are beginning to tackle this. The
greatest advances have been made in facial recognition, so that it is now
possible to automatically name individuals in a picture. Indexing is one of
the ways of improving retrieval performance. However, the costs of indexing
are high and there are continued efforts to find ways of automating this
process. The development of machine learning systems to analyse and
categorise text-based works is already well advanced. However, some text,
particularly manuscripts, depend on human interpretation. People are still
required to set up classification systems or subject terminologies that reflect
the areas being indexed. Some machine learning systems do not use
controlled vocabulary, but work on the basis of feedback from users to build
up a ‘picture’ of what the user wants. These heuristic systems are based on
experience and not necessarily codified into any formal rules.
Purpose 3: Managing information resources (Chapter 7)

Management of information resources continues to be a major factor in the
application of metadata. Since the first edition this purpose has expanded to
encompass management of research data, as well as linked data applications.
Metadata in the office environment has to deal with collections of
unstructured documents in a variety of formats. Each document may have its
own internal structure, but there is not necessarily any consistency of
document structure across the whole collection. The concept of the document
or digital lifecycle is still a helpful means of identifying the different activities
and procedures that are required during a document’s life in a collection.
Chapter 7 showed how metadata can be used to manage the workflow of
records and archival materials.
One of the key challenges remains the effective and consistent management
of information. Metadata standards help this process by making the job of
application developers easier. Although standards have been created to codify
the management of records (ISO 15489), these are in very general terms and
have not had the impact on the document and records management systems
market that was anticipated in 2004. In the library and information field,
metadata standards are in transition. Although MARC 21 has become a de
facto standard for exchange of bibliographic data, the standard is due to be
replaced by BIBFRAME, which is intended to reflect FRBR and the new RDA
cataloguing standard. This represents a distinct conceptualisation of
bibliographic records away from item-by-item cataloguing and towards

cataloguing of works, with sub-records for expressions, manifestations and
items.
Another major challenge remains the lack of compatibility between
encoding systems. Even if the same metadata standard is used, the content
of individual fields (data elements) may be different.
Preservation was included in the management category, as it is one of the
stages in the lifecycle of information resources or entities described by the
metadata. The PREMIS metadata standard was developed for managing
digital resources and is geared to digital preservation activities.
Purpose 4: Managing intellectual property rights (Chapter 8)

Rights and provenance are closely tied into one another when it comes to
intellectual property rights. The creator or the organisation that paid for the
creation of content may also have copyright or performance rights associated
with the document. In the museum and archive context provenance may also
be required to determine the authenticity of an item.
The open access movement, with initiatives such as Creative Commons,
has grown in the last decade and is now a major consideration for academic
publishing (Creative Commons, 2016; University of Nottingham, 2014;
Publications Office of the European Union, 2017). Issues of ownership of
intellectual property have also caused extensive debates in the social media
world, with providers such as Facebook developing policies that explicitly
state that users do not transfer their intellectual property rights to the social
media platform when they upload content that they have created, whether
that be text or photographs, sound or moving images. The metadata attached
to those items or associated with the postings can be seen as an indication of
ownership.
Rights management has been tumultuous over the last decade, particularly
in the context of music and video recording. As the technology for exchanging
recordings online became widely available, piracy was a major preoccupation
for the recording industry. Downloading music via the internet, even when
legitimate, raised problems of paying fees and royalties based on use of
individual items. The lack of control of downloading has led to a
transformation of the music industry, with much greater emphasis on live
performance – an outcome that perhaps was not envisaged 10 or 15 years ago.
Attempts by the recording industry to prosecute music fans for uncontrolled
and technically illegal copies of music performances backfired with a few,
well publicised cases. Where listeners do use established and regulated sites
for purchasing licences for music tracks the metadata associated with those
tracks allows large numbers of small transactions to be aggregated into global

payments to artists and groups.
Purpose 5: Supporting e-commerce and e-government (Chapter 9)

E-commerce and e-government both support transactions over the internet.
E-government is targeted at the citizen and has evolved from the early
internet days of simply providing information to more interactive services
such as submission of tax returns, registration of life events and applying for
benefits.
Online behavioural advertising accounts for a very large volume of online
transactions. The industry relies on capturing metadata about the online
behaviour of individual consumers. This enables them to target advertising
at appropriate groups. OBA raises issues of privacy in light of the large
amount of personal data that is gathered from social media.
E-commerce is exemplified by the ONIX system, which is widely used by
the global book and publishing industry. Common metadata standards allow
disparate systems to communicate and exchange information and readily
allows different models of commerce to co-exist. This means that a publisher
may communicate directly with a bookshop, or may work through an
aggregator who distributes to multiple bookshops. It also allows publishers
to use several different aggregating services covering different geographical
territories.
E-government has placed a lot of emphasis on communication of
information to citizens and residents. This has led to the development of
metadata standards and encoding schemes for governmental information.
More recently there has been a move to make government data sources
available for commercial exploitation through the open data movement.
Individual agencies have developed interfaces to allow individuals to do
simple transactions online, such as registration of life events, applications for
passports and submission of tax returns.
Purpose 6: Information governance (Chapter 10)

With the increases in concern about security and ongoing requirements to
comply with legislation, information governance has become more important.
Metadata plays a key role in providing evidence of transactions and an audit
trail of who has created, amended or accessed specific documents or data.
Although information governance can be seen as a sub-set of information
management, it is a sufficiently important activity to be recognised as a
purpose of metadata in its own right.
Verifying the authenticity of data is another area of growing concern. Many

organisations are moving away from archiving paper records to digital
preservation and management of electronic records. Providing an audit trail
of individual documents will become increasingly important to validate their
content and to ‘prove’ their authenticity and integrity. Electronic documents
are perceived as being susceptible to alteration after they have been finalised.
In practice they are no more susceptible than paper documents, especially if
there is metadata embedded in the document. Many common office
applications generate their own metadata, which includes a trail of alterations
made to the document throughout its life. An obvious step would be to
standardise this type of metadata. Because legal admissibility and evidential
weight of electronic documents depends on jurisdiction, there are national
standards such as that produced by the BSI (2014).
Managing metadata itself

The third part of this book is largely new material that has arisen from a
consideration of the challenges of managing metadata itself. This book set out
to introduce the concept of metadata and the ideas that led to its development.
This provided a basis for a description of the purposes of metadata. It is
helpful to consider this from two perspectives: that of a manager of
information resources (whether that be a librarian, archivist, museum curator,
digital repository manager or web manager); and that of user. We all consume
information for leisure, for work, for survival in a digital economy. A lot of
information is thrown at us via social media and through online behavioural
advertising, through broadcast media, and by government. We also seek out
information in response to specific needs or out of curiosity. It may be a
clearly defined need that can be addressed by a simple search. Often it is not
straightforward. We may have difficulty describing our requirements or we
may be unfamiliar with the language that is used by those that provide the
information we are seeking. I hope that this book has opened up some of the
possibilities and has provided a better understanding of the way in which
metadata contributes to this.
It is clear that it is necessary to control the content of metadata fields if there
is to be consistent performance and retrieval. The development of encoding
schemes has evolved into the creation of ontologies to represent specific areas
of knowledge. These allow for more complex relationships than the simple
hierarchies that are usually found in classification schemes. The growth of
the internet has stimulated the development of thinking about faceted
classification and groups such as the International Society for Knowledge
Organization have organised conferences and seminars to explore the
development of faceted classification (ISKO, 2017). The emergence of big data

has brought its own challenges and the use of linked data to create new
services from multiple sources of information has been an interesting
development. This seems likely to continue.
Finally, we have considered the politics of metadata. Who owns the
metadata, how is it manipulated and what are the consequences of doing so?
We have considered areas that are likely to become increasingly important in
the coming years as we see the change in political climate and the emergence
of a more authoritarian and conformist mind-set among politicians and (it
might be argued) the electorate. This raises issues of tracking use of
information, privacy and human rights. This is particularly worrying where
individual online behaviour is tracked in order to identify dissidents. At what
point is dissent considered a threat? Does expressing a view, no matter how
abhorrent to the majority of the population, constitute a crime, or should
people have the right to hold and express extreme views? These are open
questions where the debate will continue.
Conclusion
In the first edition of this book, I speculated on whether metadata as a concept
was here to stay. Metadata has been around for at least 2500 years in the form
of library catalogues. During that time it has been transformed into something
with a wide range of applications and operating in very complex
environments. Since 2004 it has become established as a label for job titles and
it has also emerged into the public awareness following high-profile news
stories. It is not a passing fad, but nor has it become a single discipline with
a coherent body of knowledge. The usage of the term ‘metadata’ is still very
varied and cuts across a number of distinct professional communities.
If there is a message from all of this, it is that the purposes of metadata
continue to be relevant and provide a useful insight into the way in which
metadata standards operate. Recent events have also demonstrated the wide
relevance of metadata to the everyday activity of all of us as users of the
internet or of telecommunications networks. The data and metadata
generated by our activity is a powerful marketing and monitoring tool that
can enhance and enrich our lives. However, we have to be aware of the harm
that can result from misuse and be vigilant about encroachments on privacy
and human rights.
References
Abele, A., McCrae, J. P., Buitelaar, P., Jentzsch, A., Cyganiak, R. (2017) The Linking
Open Data Cloud Diagram, http://lod-cloud.net [accessed 24 March 2017].
Aitchison, J., Gilchrist, A. and Bawden, D. (2000) Thesaurus Construction and Use: a
practical manual, 4th edn, London, Aslib/IMI.
Anderson, R. D. and Eberlein, K. J. (2015) Darwin Information Typing Architecture
(DITA) Version 1.3, Part 0: Overview, 10, http://docs.oasis-open.org/dita/dita/
v1.3/os/part0-overview/dita-v1.3-os-part0-overview.pdf [accessed 6 March 2017].
Arms, W. Y., Hillmann, D., Lagoze, C., Krafft, D., Marisa, R., Saylor, J., Terrizzi, C.
and Van de Sompel, H. (2002) A Spectrum of Interoperability: the site for science
prototype for the NSDL, D-Lib Magazine, 8 (1).
ASIS&T (2017) Classification Research SIG, www.asist.org/groups/classification-
research-cr [accessed 2 February 2017].
Auer, S., Bryl, V. and Tramp, S. (eds) (2014) Linked Open Data: creating knowledge out
of interlinked data. Results of the LOD2 project, Springer International Publishing.
Australian Institute of Health and Welfare (2017) METeOR,
http://meteor.aihw.gov.au/content/index.phtml/itemId/181162 [accessed 18
January 2017].
Baca, M. (ed.) (1998) Introduction to Metadata: pathways to digital information, 1st edn,
Los Angeles, CA, Getty Information Institute.
Baca, M. (ed.) (2016) Introduction to Metadata, 3rd edn,
www.getty.edu/publications/intrometadata [accessed 4 November 2016].
Baeza-Yates, R. and Ribeiro-Neto, B. (2011) Modern Information Retrieval: the concepts
and technology behind search, 2nd edn, Harlow, Pearson Education.
Bailie, R. A. and Urbina, N. (2012) Content Strategy: connecting the dots between
business, brand, and benefits, XML Press.
Baker, T., Coyle, K. and Petiya, S. (2014) Multi-entity Models of Resource
Description in the Semantic Web: a comparison of FRBR, RDA and BIBFRAME,
Library Hi Tech, 32 (4), 562–82.
240 METADATA FOR INFORMATION MANAGEMENT AND RETRIEVAL
Basel University Library (2017) Basel Register of Thesauri, Ontologies and

Classifications, https://bartoc.org [accessed 2 February 2017].
Bayes, T. and Price, R. (1763). An Essay Towards Solving a Problem in the Doctrine
of Chance, Philosophical Transactions of the Royal Society of London, 53, 370–418.
BBC (2017) Domesday Reloaded, www.bbc.co.uk/history/domesday/story [accessed
25 May 2017].
Bell, G. (2016) Interview with Graham Bell.
Bernardi, R., Cakici, R., Elliott, D., Erdem, A., Erdem, E., Ikizler-Cinbis, N., Keller,
F., Muscat, A. and Plank, B. (2016) Automatic Description Generation from
Images: a survey of models, datasets, and evaluation measures, Journal of
Artificial Intelligence Research, 55, 409–42.
Berners-Lee, T. (1998) Semantic Web Roadmap,
www.w3.org/DesignIssues/Semantic.html [accessed 12 April 2017].
Berners-Lee, T., Fielding, R. and Masinter, L. (2005) Uniform Resource Identifier (URI):
generic syntax, 61, https://tools.ietf.org/pdf/rfc3986.pdf [accessed 19 August 2015].
Bhatnagar, S. (2009) Unlocking E-government Potential: concepts, cases and practical
insights, New Delhi, SAGE Publications.
Bholat, D., Hansen, S., Santos, P. and Schonhardt-Bailey, C. (2015) Text Mining for
Central Banks, CCBS Handbook No. 33, London.
Bide, M. (2011) Identifier and Metadata Standards for E-Commerce: responding to
reality in 2011, Journal of Electronic Publishing, 14 (1).
Blackburn, B., Smallwood, R. and Earley, S. (2014) Information Organization and
Classification: taxonomies and metadata. In Smallwood, R. F. (ed.) Information
Governance: concepts, strategies and best practices, Hoboken, NJ, John Wiley and
Sons, 355–84.
Bond, R. (2010) Data Ownership in Social Networks – a very personal thing, Privacy
and Data Protection, 11 (1), 1–5.
Bovalis, K., Peristeras, V., Abecasis, M., Abril-Jimenez, R., Rodriguez, M. A.,
Gattegno, C., Karalopoulos, A., Sagias, I., Szekacs, S. and Wigard, S. (2014)
Promoting Interoperability in Europe’s E-Government, Computer, 47 (1), 25–33.
Bowman, J. H. (2003) Essential Cataloguing: the basics, London, Facet Publishing.
Bradley, P. (2013) Expert Internet Searching, 4th edn, London, Facet Publishing.
Bray, T., Hollander, D., Layman, A., Tobin, R. and Thompson, H. S. (2009)
Namespaces in XML, www.w3.org/TR/REC-xml-names [accessed 19 April 2017].
Breeding, M., Kroeger, A. and Sandy, H. M. (2016) Sharing Metadata across
Discovery Systems. In Spiteri, L. F. (ed.) Managing Metadata in Web-Scale Discovery
Systems, London, Facet Publishing, 17–55.
Brickley, D. and Miller, L. (2014) FOAF Vocabulary Specification 0.99,
http://xmlns.com/foaf/spec/20140114.html [accessed 27 July 2015].
Brighton, J. (2011) Introducing PBCore 2.0, http://pbcore.org/news/introducing-
pbcore-2-0 [accessed 27 July 2015].
REFERENCES 241
British Library Labs (2017) LibCrowds, www.libcrowds.com [accessed 24 March 2017].

Broughton, V. (2006) Essential Thesaurus Construction, London, Facet Publishing.
Broughton, V. (2015) Essential Classification, 2nd edn, London, Facet Publishing.
BSI (2011) BS 0:2011 A Standard for Standards – Principles of standardization, London.
BSI (2014) Evidential Weight and Legal Admissibility of Electronic Information –
Specification, London.
Buckland, M. K. (1997) What is a Document?, Journal of the American Society for
Information Science, 48 (9), 804–9.
Caldwell, B., Cooper, M., Reid, L. G. and Vanderheiden, G. (2008) Web Content
Accessibility Guidelines (WCAG) 2.0, W3C, www.w3.org/TR/WCAG20 [accessed
23 March 2017].
Caplan, P. (2003) Metadata Fundamentals for all Librarians, Chicago, IL, American
Library Association.
CCSDS (2012) Reference Model for an Open Archival Information System (OAIS),
Recommended Practice CCSDS 650.0-M-2, Washington, DC.
Central Information Office (2016) Project Open Data Metadata Schema ver. 1.1,
Washington, DC.
Ceri, S., Bozzon, A., Brambilla, M., Della Valle, E., Fraternali, P. and Quarteroni, S.
(2013) Web Information Retrieval, Berlin, Springer Verlag.
Chaffey, D. (2015) Digital Business and E-commerce Management: strategy,
implementation and practice, 6th edn, Harlow, Pearson Education.
Chan, L. M. and Salaba, A. (2016) Cataloging and Classification: an introduction, 4th
edn, Lanham, MD, Rowman and Littlefield Publishers.
Chen, X. and Li, G. (2013) Evaluation Indicators and Model of Network Technical
Anonymity, International Journal of Future Generation Communication and
Networking, 6 (4), 181–92.
Chen, Y.-N., Chen, S.-J. and Lin, S.C. (2003) A Metadata Lifecycle Model for Digital
Libraries: methodology and application for an evidence-based approach to
library research. In IFLA Conference Proceedings, Berlin, 1–15.
Cloud Foundry Foundation (2017) Catalog Metadata,
https://docs.cloudfoundry.org/services/catalog-metadata.html [accessed 3 March
2017].
Cochrane, P. A. (1982) Subject Access in the Online Catalog, Dublin, OH., OCLC.
Cole, T., Habing, T., Hunter, J., Johnston, P. and Lagoze, C. (2008) DCMES 1.1 XML
Schema, http://dublincore.org/schemas/xmls/qdc/2008/02/11/dc.xsd [accessed 24
April 2017].
Cole, T. W. and Han, M.-J. K. (2013) XML for Catalogers and Metadata Librarians, Santa
Barbara, CA, Libraries Unlimited.
Colvin, E. and Kraft, D. H. (2016) Fuzzy Retrieval for Software Reuse, Journal of the
Association for Information Science and Technology, 67 (10), 2454–63.
Corcho, O., Poveda-Villalón, M. and Gómez-Pérez, A. (2015) Ontology Engineering
in the Era of Linked Data, Bulletin of the Association for Information Science and
Technology, 41 (4), 13–17.
Corporation for Public Broadcasting (2011) PBCore Schema,
http://pbcore.org/schema/ [accessed 27 July 2015].
Coyle, K. and Baker, T. (2009) Guidelines for Dublin Core Application Profiles,
http://dublincore.org/documents/2009/05/18/profile-guidelines [accessed 21 July
2015].
Creative Commons (2016) Creative Commons, http://creativecommons.org [accessed 5
February 2016].
Daconta, M. C. (2013) Big Metadata: 7 ways to leverage your data in the cloud, GCN,
https://gcn.com/blogs/reality-check/2013/12/metadata.aspx [accessed 3 March
2017].
DataCite Metadata Working Group (2016) DataCite Metadata Schema Documentation
for the Publication and Citation of Research Data, Version 4.0,
http://schema.datacite.org/meta/kernel-4.0/doc/DataCite-
MetadataKernel_v4.0.pdf [accessed 2 October 2017].
Davenport, T. H. (2014) Big Data at Work: dispelling the myths, uncovering the
opportunities, Boston, MA, Harvard Business Review Press.
Day, M. (1999) Metadata for Digital Preservation: an update, Ariadne, 22,
www.ariadne.ac.uk/issue22/metadata [accessed 4 October 2017].
Day, M. (2001) Metadata in a Nutshell, Information Europe, 6 (2), 11.
Day, M. (2004) Preservation Metadata. In Gorman, G. E. (ed.) Information Yearbook for
Library and Information Management 2003–2004: metadata applications and
management, London, Facet Publishing, 253–73.
DCMI (2012) Dublin Core Metadata Element Set v1.1,
http://dublincore.org/documents/dces [accessed 24 April 24 2017].
DCMI (2015) DCMI Mission and Principles, http://dublincore.org/about-us [accessed
21 July 2015].
DCMI Usage Board (2012) DCMI Metadata Terms,
http://dublincore.org/documents/2012/06/14/dcmi-terms [accessed 18 August
2015].
de Vries, J., Williams, T. N., Bojang, K., Kwiatkowski, D. P., Fitzpatrick, R. and
Parker, M. (2014) Knowing Who to Trust: exploring the role of ‘ethical metadata’
in mediating risk of harm in collaborative genomics research in Africa, BMC
Medical Ethics, 15 (62), http://www.biomedcentral.com/1472-6939/15/62 [accessed
2 October 2017]
de Waard, A. (2016) Research Data Management at Elsevier: supporting networks of
data and workflows, Information Services & Use, 36 (1–2), 49–55.
Democracy Now (2013) Court: Gov’t Can Secretly Obtain Email, Twitter Info from
Ex-WikiLeaks Volunteer Jacob Appelbaum,
www.democracynow.org/2013/2/5/court_govt_can_secretly_obtain_email
REFERENCES 243
[accessed 21 March 2017].

Detlor, B. (2010) Information Management, International Journal of Information
Management, 30 (2), 103–8.
Dewey, C. (2016) 98 Personal Data Points that Facebook Uses to Target Ads to You,
Washington Post, 19 August 2016.
Digital Curation Centre (2010) DCC Curation Lifecycle Model,
www.dcc.ac.uk/resources/curation-lifecycle-model [accessed 6 January 2016].
Digital Curation Centre (2017) Disciplinary Metadata,
www.dcc.ac.uk/resources/metadata-standards [accessed 17 January 2017].
Drexel University (2017) Metadata Research Center, https://cci.drexel.edu/mrc
[accessed 31 January 2017].
Duff, A. S. (1997) Some Post-War Models of the Information Chain, Journal of
Librarianship and Information Science, 29 (4), 179–87.
Duranti, L. (1989) Diplomatics: new uses for an old science, Archivaria, 28, 7–27.
Duranti, L. and Rogers, C. (2012) Trust in Digital Records: an increasingly cloudy
legal area, Computer Law & Security Review, 28 (5), 522–31.
e-Government Unit (2006) e-Government Metadata Standard ver. 3.1,
www.nationalarchives.gov.uk/documents/information-management/
egms-metadata-standard.pdf [accessed 24 March 2016].
Earley, S. (2017) Earley Information Science – resource center,
www.earley.com/resource-center [accessed 31 January 2017].
EBU (2015) TECH 3293 EBU Core Metadata Set (EBUCore) Specification v1.6, 40,
https://tech.ebu.ch/docs/tech/tech3293.pdf [accessed 27 July 2015].
EDItEUR (2014) ONIX for Books. Implementation and best practice guide, Release 3.0
rev. 2, www.editeur.org/93/Release-3.0-Downloads/#Best practice [accessed 19
August 2015].
Europeana Labs (2016) Europeana Data Model. Data structure,
http://labs.europeana.eu/api/linked-open-data-data-structure [accessed 24 March
2017].
European Commission (2011) Towards Open Government Metadata,
https://joinup.ec.europa.eu/sites/default/files/24/4c/14/towards_open_government_
metadata_0.pdf [accessed 3 November 2016].
European Commission (2017a) European Data Portal, www.europeandataportal.eu/
[accessed 3 March 2017].
European Commission (2017b) Technical Guidelines for Implementing Dataset and
Service Metadata based on ISO/TS 19139:2007,
http://inspire.ec.europa.eu/id/document/tg/metadata-iso19139/2.0.1, [accessed 2
October 2017].
European Parliament and European Council (2016) General Data Protection Regulation
– EU 2016/679, EU: OJ L 119 04.05.2016, 1–88.
Eysenbach, G., Kohler, C., Yihune, G., Lampe, K., Cross, P. and Brickley, D. (2001) A
Metadata Vocabulary for Self- and Third-party Labeling of Health Web-sites:

Health Information Disclosure, Description and Evaluation Language (HIDDEL),
Annual Symposium of the American-Medical-Informatics-Association (AMIA 2001)
Location: Washington, D.C., 3-7 November 2001.
Facebook (2014) Open Graph Protocol, http://ogp.me [accessed 27 July 2015].
Farias Lóscio, B., Burle, C. and Calegari, N. (2017) Data on the Web Best Practices,
www.w3.org/TR/dwbp [accessed 3 March 2017].
Feather, J. and Sturges, P. (eds) (1997) International Encylcopedia of Information and
Library Science, London, Routledge.
Federal CIO Council (2017) Project Open Data Metadata Resources,
https://project-open-data.cio.gov/metadata-resources [accessed 16 February
2017].
Field, T., Muller, E. and Lau, E. (2003) The E-Government Imperative, Paris, OECD.
Floridi, L. (2010) Information: a very short introduction, Oxford, Oxford University
Press.
Galeffi, A., Bertolini, M. V., Bothmann, R. L., Rodríguez, E. E., McGarry, D. (2016)
Statement of International Cataloguing Principles (ICP), The Hague, IFLA.
Galloway, D. F. (1958) Machine-Tools. In Singer, C. et al. (eds) A History of
Technology, Volume V – The Late Nineteenth Century, c1850–c1900, Oxford,
Clarendon Press, 636–57.
Gartner, R. (2016) Metadata: Shaping Knowledge from Antiquity to the Semantic Web,
Cham, Switzerland, Springer International Publishing.
Getty Images (2017) Getty Images, www.gettyimages.co.uk [accessed 20 April 2017].
Gilliland, A. J. (2016) Setting the Stage. In Baca, M. (ed.) Introduction to Metadata, Los
Angeles, CA, Getty Publications.
Goldacre, B. (2013) Bad Pharma: how medicine is broken, and how we can fix it, London,
Fourth Estate.
Goldfarb, C. F. (1990) The SGML Handbook, Oxford, Clarendon Press.
Goldfarb, C. F. and Prescod, P. (2001) The XML Handbook, 3rd edn, Upper Saddle
River, NJ, Prentice Hall.
Golub, K., Soergel, D., Buchanan, G., Tudhope, D., Lykke, M. and Hiom, D. (2016) A
Framework for Evaluating Automatic Indexing or Classification in the Context of
Retrieval, Journal of the Association for Information Science and Technology, 67 (1),
3–16.
Gorman, M. (2004) Authority Control in the Context of Bibliographic Control in the
Electronic Environment. Cataloging and Classification Quarterly, 38 (3/4), 11–22.
Greenberg, J. (2009) Theoretical Considerations of Lifecycle Modeling: an analysis of
the Dryad repository demonstrating automatic metadata propagation,
inheritance, and value system adoption, Cataloging & Classification Quarterly, 47,
380–402.
Greenwald, G. (2013) NSA Collecting Phone Records of Millions of Verizon
REFERENCES 245
Customers Daily, Guardian, 6 June 2013.

Greenwald, G. (2014) No Place to Hide, London, Hamish Hamilton.
Gruber, T. (2009) Ontology. In Liu, L. and Özsu, M. T. (eds) Database Systems,
Springer Verlag.
Grunzke, R., Mueller-Pfefferkorn, R., Jaekel, R., Starek, J., Hardt, M., Hartmann, V.,
Potthoff, J., Hesser, J., Kepper, N., Gesing, S. and Kindermann, S. (2014) Device-
driven Metadata Management Solutions for Scientific Big Data Use Cases. In
Aldinucci, P., DAgostino, M. and Kilpatrick, D. (eds) 22nd EuroMicro International
Conference on Parallel, Distributed, and Network-based Processing (PDP 2014), New
York, NY, IEEE, 317–21.
GS1 (2015) GS1 General Specifications, Version 15, Brussels.
Gueguen, G., Marques da Fonseca, V. M., Pitti, D. V. and de Grimoüard, C. S. (2013)
Toward an International Conceptual Model for Archival Description: a
preliminary report from the International Council on Archives’ Experts Group on
Archival Description, American Archivist, 76 (2), 566–683.
Guy, M. and Tonkin, E. (2006) Folksonomies: tidying up tags?, D-Lib Magazine, 12
(1).
Hansen, D. L., Schneiderman, B. and Smith, M. A. (2011) Analyzing Social Media
Networks with NodeXL: insights from a connected world, Burlington, MA, Morgan
Kaufmann Publishers.
Hansen, J. and Andresen, L. (2003) Dublin Core DCMI Administrative Metadata. Final
Version, www.bs.dk/standards/AdministrativeComponents.htm [accessed 20
January 2017].
Harpring, P. (2014) Metadata Standards Crosswalk, J. Paul Getty Trust,
www.getty.edu/research/conducting_research/standards/intrometadata/
crosswalks.html [accessed 19 January 2017].
Hay, D. C. (2006) Data Model Patterns: a metadata map, San Francisco, CA, Morgan
Kaufmann Publishers.
Haynes, D. (2004) Metadata for Information Management and Retrieval, 1st edn,
London, Facet Publishing.
Haynes, D., Huckle, F. and Elliot, H. (2002) HSE Thesaurus, Bootle.
Hedden, H. (2016) The Accidental Taxonomist, Medford, NJ, Information Today.
Herman, I., Adida, B., Sporny, M. and Birbeck, M. (2015) RDFa 1.1 Primer, 3rd edn,
W3C, www.w3.org/TR/rdfa-primer [accessed 26 April 2017].
Hider, P. (2012) Information Resource Description: creating and managing metadata,
Hillmann, D., Dunsire, G. and Phipps, J. (2013) Maps and Gaps: strategies for
vocabulary design and development. In International Conference on Dublin Core
and Metadata Applications. Lisbon, 82–89. Foulonneau, M. and Eckert,K (editors),
Dublin OH, Dublin Core Metadata Initiative,
http://dcpapers.dublincore.org/pubs/article/view/3673/1896 [accessed 2 October,
2017].
Hirata, K. and Kato, T. (1992) Query by Visual Example. In Pirotte, A., Delobel, C.
and Gottlob, G. (eds) Advances in Database Technology – EDBT ’92: 3rd International
Conference on Extending Database Technology Vienna, Austria, March 23–27, 1992
Proceedings, Berlin, Heidelberg, Springer, 56–71.
Hudgins, J., Agnew, G. and Brown, E. (1999) Getting Mileage out of Metadata:
applications for the library, LITA Guides No. 5, Chicago, IL, American Library
Association.
Iannella, R. (2002) Open Digital Rights Language (ODRL) Version 1.1, Cambridge MA,
W3C, https://www.w3.org/TR/2002/NOTE-odrl-20020919/ [accessed 2 October
2017].
Iannella, R. and Campbell, D. (1999) The A-Core: metadata about content metadata,
http://metadata.net/admin/draft-iannella-admin-01.txt [accessed 2 March 2004].
ICA (2000) ISAD(G): General International Standard Archival Description, 2nd edn,
www.icacds.org.uk/eng/ISAD(G).pdf, 91 [accessed 12 April 2016].
ICOT (1999) What is FGCS Technologies?,
https://web.archive.org/web/20090217105259/www.icot.or.jp/ARCHIVE/
Museum/ICOT/FGCS-E.html [accessed 24 March 2017].
IEEE Computer Society (2002) 1484.12.1 IEEE Standard for Learning Object Metadata,
New York, NY, IEEE.
IFFF Consortium (2017) IIIF Presentation API 2.1.1.,
http://iiif.io/api/presentation/2.1/#b-summary-of-metadata-requirements
[accessed 17 August 2017].
IFLA (1998) Functional Requirements for Bibliographic Records Final Report: IFLA Study
Group on the Functional Requirements for Bibliographic Records, Munich, K. G. Saur.
IFLA (2011) ISBD – International Standard Bibliographic Description Consolidated
Edition, Berlin, De Gruyter Saur.
IFLA (2013) Functional Requirements for Authority Data: a conceptual model, The Hague.
Information Commissioner’s Office (2012) Anonymisation: managing data protection
risk code of practice, Wilmslow.
International Council on Archives (2004) ISAAR (CPF) International Standard Archival
Authority Record for Corporate Bodies, Persons and Families, Paris.
International DOI Foundation (2012) DOI Handbook, www.doi.org/hb.html [accessed
31 March 2016].
International ISBN Agency (2012) ISBN Users’ Manual, 6th edn, London.
International ISBN Agency (2014) What is an ISBN?,
www.isbn-international.org/content/what-isbn [accessed 8 May 2017].
International ISTC Agency (2010) International Standard Text Code (ISTC) User
Manual, Version 1.2, London.
IPTC (2014) IPTC Photo Metadata Standard, London.
ISKO (2017) International Society for Knowledge Organization, www.isko.org [accessed
REFERENCES 247
2 February 2017].
ISO (1986) ISO 8879:1986 Information Processing – Text and Office Systems – Standard
Generalized Markup Language (SGML), Geneva.
ISO (2002) ISO 639-1:2002 Codes for the Representation of Names of Languages – Part 1:
Alpha-2 Code, Geneva.
ISO (2004a) ISO/IEC 15444-2:2004 Information Technology – JPEG 2000 Image Coding
System – Part 2: Extensions, Geneva.
ISO (2004b) ISO/IEC TR 21000-1:2004 Information T – Multimedia Framework (MPEG-
21) – Part 1: Vision, Technologies and Strategy, Geneva.
ISO (2004c) ISO 8601:2004 Data Elements and Interchange Formats – Information
Interchange – Representation of Dates and Times, Geneva.
ISO (2005) ISO 2108:2005 Information and Documentation – International Standard Book
Number (ISBN), Geneva.
ISO (2007) ISO 3297:2007 Information and Documentation - International Standard Serial
Number (ISSN), Geneva.
ISO (2008) ISO15706-1:2002+A1:2008 Information and Documentation – International
Standard Audiovisual Number (ISAN), Geneva.
ISO (2009a) ISO 10957:2009 Information and Documentation. International Standard
Music Number (ISMN), Geneva.
ISO (2009b) ISO 15836:2009 Information and Documentation – The Dublin Core Metadata
Element Set, Geneva.
ISO (2009c) ISO 21047:2009 Information and Documentation – International Standard
Text Code (ISTC), Geneva.
ISO (2009d) ISO 23081-2:2009 Information and Documentation – Records Management
Processes – Metadata for Records – Part 2: Conceptual and Implementation Issues,
Geneva.
ISO (2009e) ISO 31000:2009 Risk Management – Principles and Guidelines, Geneva.
ISO (2011) ISO 25964-1:2011 – Information and Documentation – Thesauri and
Interoperability with other Vocabularies. Part 1: Thesauri for Information Retrieval,
Geneva.
ISO (2012a) ISO/IEC 19505-1: 2012 Information technology – Object Management Group
Unified Modeling Language (OMG UML) – Part 1: Infrastructure, Geneva.
ISO (2012b) ISO 26324:2012 Information and Documentation – Digital Object Identifier
System, Geneva.
ISO (2013) ISO 3166-1:2013 Codes for the Representation of Names of Countries and Their
Subdivisions Part 1: Country Codes, Geneva.
ISO (2014a) EN ISO 19115-1 Geographic information – Metadata. Part 1: Fundamentals,
Geneva.
ISO (2014b) ISO 28560-2:2014 Information and Documentation – RFID in Libraries – Part
2: Encoding of RFID Data Elements Based on Rules from ISO/IEC 15962, Geneva.
ISO (2015) Standards, www.iso.org/iso/home/standards.htm [accessed 6 July 2015].
ISO (2016a) ISO 15489-1:2016 Information and Documentation, Records Management,

Concepts and Principles, Geneva.
ISO (2016b) ISO 24617-6:2016 Language Resource Management – Semantic Annotation
Framework – Part 6: Principles of Semantic Annotation (SemAF Principles), Geneva.
ISO/IEC (2015) ISO/IEC 11179-1:2015(E) Information Technology — Metadata Registries
(MDR) – Part 1: Framework, Geneva.
ISSN International Centre (2017) International Standard Serial Number, www.issn.org
[accessed 8 May 2017].
iStockphoto LP (2017) iStock, www.istockphoto.com [accessed 20 April 2017].
ITU-T (2014) ITU-T Rec. X.667 (10/2012) Information Technology – Procedures for the
Operation of Object Identifier Registration Authorities: Generation of Universally
Unique Identifiers and Their Use in Object Identifiers, Geneva.
Jansen, A. (2014) Preservation as a Service for Trust: an InterPARES trust
specification for preserving authentic records in the cloud. In Endicott-Popovsky,
B. (ed.) Proceedings of the International Conference on Cloud Security Management
(ICCSM-2014), 67–72.
Joint Steering Committee for Development of RDA (2014) Resource Description and
Access: RDA – 2014 Revision, London, CILIP.
Joint Steering Committee for Development of RDA (2015) MARC Bibliographic to
RDA Mapping, http://access.rdatoolkit.org/jscmap2.html [accessed 22July 2015].
Joint Steering Committee for Revision of AACR & CILIP (2002) Anglo-American
Cataloguing Rules (AACR2) 2nd ed., Ottawa; London; Chicago IL: Canadian
Library Association; CILIP; American Library Association.
Karpathy, A. and Fei-Fei, L. (2017) Deep Visual-Semantic Alignments for Generating
Image Descriptions, IEEE Transactions on Pattern Analysis and Machine Intelligence,
39 (4), 664–76.
Kerrigan, S. and Law, K. (2005) Regulation-Centric, Logic-Based Compliance
Assistance Framework, Journal of Computing in Civil Engineering, 19 (1), 1–15.
Khanuja, H. and Suratkar, S. S. (2014) Role of Metadata in Forensic Analysis of
Database Attacks. In 2014 IEEE International Advance Computing Conference
(IACC), 457–62.
Kitchin, R. (2014) The Data Revolution: big data, open data, data infrastructures and their
consequences, London, SAGE Publications.
Konkova, E., Göker, A., Butterworth, R. and MacFarlane, A. (2014) Social Tagging:
exploring the image, the tags, and the game, Knowledge Organization, 41 (1), 57–65.
Kunze, J. A. (2003) Towards Electronic Persistence Using ARK Identifiers, Oakland, CA.
Available at https://confluence.ucop.edu/download/attachments/
16744455/arkcdl.pdf [accessed June 1, 2017].
Lagoze, C., Van de Sompel, H., Nelson, M. and Warner, S. (2002) The Open Archives
Initiative Protocol for Metadata Harvesting,
www.openarchives.org/OAI/openarchivesprotocol.html [accessed 18 June 2015].
REFERENCES 249
Lagoze, C. and Hunter, J. (2002) The ABC Ontology and Model, Journal of Digital
Information, 2 (2), https://journals.tdl.org/jodi/index.php/jodi/article/view/44/47
[accessed 2 October. 2017].
Lambe, P. (2007) Organising Knowledge, Oxford, Chandos Publishing.
LaTeX3 Project Team (2001) LaTeX2e for Authors, https://www.latex-
project.org/help/documentation/usrguide.pdf [accessed 2 October 2017].
Latham, K. F. (2012) Museum Object as Document: using Buckland’s information
concepts to understand museum experiences, Journal of Documentation, 68 (1),
45–71.
Laudon, K. C. and Traver, C. G. (2014) E-commerce: business, technology, society,
Global edn, 10th edn, Harlow, Pearson Education.
Lavoie, B. F. (2004) The Open Archival Information System Reference Model: introductory
guide, Dublin, OH, OCLC.
Law, K. H., Lau, G., Kerrigan, S. and Ekstrom, J. A. (2014) REGNET: regulatory
information management, compliance and analysis, Government Information
Quarterly, 31, S37–S48.
Li, C. and Sugimoto, S. (2014) Provenance Description of Metadata using PROV
with PREMIS for Long-term Use of Metadata. In International Conference on
Dublin Core and Metadata Applications, 8–11 October 2014, Austin, TX, 147–56.
Library of Congress (2011) MIX NISO Metadata for Images in XML Schema,
www.loc.gov/standards/mix [accessed 27 July 2015].
Library of Congress (2012) Dublin Core Metadata Element Set Mapping to MODS
Version 3, www.loc.gov/standards/mods/dcsimple-mods.html [accessed 19
January 2017].
Library of Congress (2015a) Metadata Encoding and Transmission Standard (METS)
Official Web Site, www.loc.gov/standards/mets [accessed 29 July 2015].
Library of Congress (2015b) Metadata Object Description Schema (MODS),
www.loc.gov/standards/mods [accessed 23 July 2015].
Library of Congress (2016a) Encoded Archival Description (EAD),
www.loc.gov/ead [accessed 24 March 2016].
Library of Congress (2016b) External Schemas for Use with METS,
www.loc.gov/standards/mets/mets-extenders.html [accessed 26 May 2017].
Library of Congress (2017a) Bibliographic Framework Initiative (BIBFRAME),
www.loc.gov/bibframe [accessed 17 August 2017].
Library of Congress (2017b) Library of Congress Names,
http://id.loc.gov/authorities/names.html [accessed 8 May 2017].
Library of Congress (2017c) MARC Standards, www.loc.gov/marc [accessed 28
March 2017].
Library of Congress and Stock Artists Alliance (2009) PhotoMetadata Project,
www.photometadata.org [accessed 24 September 2015].
MacFarlane, A., Robertson, S. E. and McCann, J. A. (2004) Parallel Computing for
Passage Retrieval, Aslib Proceedings, 56 (4), 201–11.

Malta, M. C. and Baptista, A. A. (2014) A Panoramic View on Metadata Application
Profiles of the Last Decade, International Journal of Metadata, Semantics and
Ontologies, 9 (1), 58–73.
Matthews, B., Jones, C., Puzoń, B., Moon, J., Tudhope, D., Golub, K. and Nielsen, M.
L. (2010) An Evaluation of Enhancing Social Tagging with a Knowledge
Organization System, Aslib Proceedings, 62 (4/5), 447–65.
Mayer-Schönberger, V. and Cukier, K. (2013) Big Data: a revolution that will transform
how we live, work and think, London, John Murray.
McClelland, M., McArthur, D., Giersch, S. and Geisler, G. (2002) Challenges for
Service Providers when Importing Metadata in Digital Libraries, D-Lib Magazine,
8 (4), http://www.dlib.org/dlib/april02/mcclelland/04mcclelland.html [accessed 2
October 2017].
McGuinness, D. L. (2002) Ontologies Come of Age. In Fensel, D. (ed.) Spinning the
Semantic Web: bringing the world wide web to its full potential, Cambridge, MA, MIT
Press.
Metadata Management Associates (2017) Open Metadata Registry,
http://metadataregistry.org [accessed 18 January 2017].
Metadata Working Group (2010) Guidelines for Handling Image Metadata. Version 2.0,
http://www.metadataworkinggroup.org/pdf/mwg_guidance.pdf [accessed 2
October 2017].
Microsoft (2017) CaptionBot, www.captionbot.ai [accessed 24 May 2017].
Miles, A. and Bechhofer, S. (2009) SKOS Simple Knowledge Organization System
Namespace Document – HTML Variant,
www.w3.org/2009/08/skos-reference/skos.html [accessed 17 January 2017].
Miranda, S. and Ritrovato, P. (2014) Automatic Extraction of Metadata from
Learning Objects. In Xhofa, F., Barolli, L., Palmieri, F., Koeppen, M. and Loia, V.
(eds), 2014 International Conference on Intelligent Networking and Collaborative
Systems, 10-12 September 2014, Salerno, Italy, New York, NY, IEEE, 704–9.
Mislove, A., Marcon, M., Gummadi, K. R., Druschel, P. and Bhattacharjee, B. (2007)
Measurement and Analysis of Online Social Networks, In IMC’07: Proceedings of
the 2007 ACM SIGCOMM Internet Measurement Conference, San Diego, CA, October
24–25 2007, New York, NY, Association for Computing Machinery, 29–42.
Moreau, L. (2010) The Foundations for Provenance on the Web, Foundations and
Trends in Web Science, 2 (2–3), 99–241.
Morrison, S. R. (2014) The System of Domestic Counterterrorism Law Enforcement,
Stanford Law and Policy Review, 25 (2), 341–77.
NAA (2010) AGLS Metadata Standard Part 1 – Reference Description, Canberra.
Narayanan, A. and Shmatikov, V. (2009) De-anonymizing Social Networks, in 2009
30th IEEE Symposium on Security and Privacy, 17–20 May, IEEE, 173–87.
National Archives of Australia (2008) AGLS Metadata Standard., www.agls.gov.au/
REFERENCES 251
[accessed March 24, 2017].

National Archives of Australia (2010) AGLS Metadata Standard Part 1 – Reference
Description, Canberra, www.agls.gov.au/pdf/AGLS Metadata Standard Part 1
Reference Description.PDF [accessed 4 October 2017].
National Institute of Standards and Technology (2017) Text REtrieval Conference
(TREC), http://trec.nist.gov [accessed 24 May 2017].
National Oceanic and Atmospheric Administration (2017) National Climatic Data
Center, www.ncdc.noaa.gov [accessed 1 March 2017].
Net7 srl, (2017) Pundit Web Annotation Data Model 1.1,
https://docs.google.com/spreadsheets/d/10XXQ5KrFZbFqKxiFhuVBQrWdSxbXd
85cCYYmrXE23QQ/edit#gid=0 [accessed 24 March 2017].
New Zealand E-Government Unit (2001) Metadata Management Facility User
Requirements Specification, www.e-government.gov.nz/docs/mmf-users/mmf-
users.pdf [accessed 13 February 2004].
Nilsson, M., Baker, T. and Johnston, P. (2008) The Singapore Framework for Dublin
Core Application Profiles, http://dublincore.org/documents/2008/01/14/
singapore-framework [accessed 3 October 2017].
Nilsson, M., Baker, T. and Johnston, P. (2009) Interoperability Levels for Dublin Core
Metadata, http://dublincore.org/documents/2009/05/01/interoperability-levels
NISO (2006) ANSI/NISO Z39.87-2006 Data Dictionary – Technical Metadata for Digital
Still Images, Bethesda, MD.
NISO/UKSG KBART Working Group (2010) KBART: knowledge bases and related tools,
Baltimore, MD.
Norris, P. (2001) Digital Divide: civic engagement, information poverty, and the internet
worldwide, Cambridge, Cambridge University Press.
OMG (2015) Unified Modelling Language, Needham, MA.
Open Archives Initiative (2017) ResourceSync Framework Specification (ANSI/NISO
Z39.99-2017), Baltimore, MD, NISO.
http://www.openarchives.org/rs/1.1/resourcesync [accessed 2 October 2017]
Open Discovery Initiative Working Group (2014) Open Discovery Initiative: promoting
transparency in discovery, Baltimore MD.
Open Knowledge International (2017) Data Portals: a comprehensive list of open data
portals from around the world, http://dataportals.org [accessed 16 February 2017].
OWL Working Group (2012) Web Ontology Language (OWL), W3C,
www.w3.org/2001/sw/wiki/OWL [accessed 19 April 2016].
Paganelli, F., Pettenati, M. and Giuli, D. (2006) A Metadata-Based Approach for
Unstructured Document Management in Organizations, Information Resources
Management Journal, 19 (1), 1–22.
Pariser, E. (2011) The Filter Bubble: what the Internet is hiding from you, London, Viking.
Park, J. and Brenza, A. (2015) Evaluation of Semi-Automatic Metadata Generation
Tools: a survey of the current state of the art, Information Technology & Libraries, 34
(3), 22–42.
Pearsall, J. (ed.) (1999) The New Oxford Dictionary of English, Oxford, Oxford
University Press.
Peters, I. and Weller, K. (2008) Tag Gardening for Folksonomy Enrichment and
Maintenance, Webology, 5 (3), http://www.webology.org/2008/v5n3/a58.html
[accessed 6 February 2017]
Pinker, S. (2015) The Sense of Style: the thinking person’s guide to writing in the 21st
century, Penguin Books.
Pomerantz, J. (2015) Metadata, Cambridge, MA, MIT Press.
Ponceleón, D. and Slaney, M. (2011) Multimedia Information Retrieval. In Baeza-
Yates, R. and Ribeiro-Neto, B. (eds) Modern Information Retrieval: the concepts and
technology behind search, Harlow, Pearson Education, 587–639.
Powell, A., Nilsson, M., Naeve, A., Johnston, P. and Baker, T. (2007) DCMI Abstract
Model, http://dublincore.org/documents/abstract-model [accessed 19 April 2016].
PREMIS Editorial Committee (2015) PREMIS Data Dictionary for Preservation
Metadata, Version 3.0, http://www.loc.gov/standards/premis/v3/
premis-3-0-final.pdf [accessed 25 May 2017].
Privacy International (2017) Privacy 101, Metadata,
www.privacyinternational.org/node/53 [accessed 23 March 2017].
Publications Office of the European Union (2017) EU Open Data Portal,
http://data.europa.eu/euodp/en/data [accessed 14 February 2017].
RDA Steering Committee (2015) Map from ISBD Properties to Unconstrained RDA
Properties, www.rdaregistry.info/Maps/mapISBD2RDAU.html [accessed 17 July
2017].
RDA Steering Committee (2017) RSC Website, www.rda-rsc.org [accessed 16 July
2017].
RDA Steering Committee, Metadata Management Associates & ALA Digital
Reference (2016) RDA Registry Examples, www.rdaregistry.info/Examples
[accessed 30 July 2017].
Riva, P., Le Boeuf, P. and Žumer, M. (2017) IFLA Library Reference Model, The Hague.
Robinson, L. (2009) Information Science: communication chain and domain analysis,
Journal of Documentation, 65 (4), 578–91.
Robinson, L. (2015) Multisensory, Pervasive, Immersive: towards a new generation
of documents, Journal of the Association for Information Science and Technology, 66
(8), 1734–7.
Rodriguez, M. A., Bollen, J. and Van de Sompel, H. (2009) Automatic Metadata
Generation Using Associative Networks, ACM Transactions on Information
Systems, 27 (2), 7:1–7:20.
Rosenberg, D. (2013) Data Before the Fact. In Gitelman, L. (ed.) ‘Raw Data’ is an
Oxymoron, Cambridge, MA, Massachusetts Institute of Technology, 15–40.
REFERENCES 253
Rousidis, D., Garoufallou, E., Balatsoukas, P. and Sicilia, M-A. (2014) Metadata for
Big Data: a preliminary investigation of metadata quality issues in research data
repositories, Information Services and Use, 34 (3–4), 279–86.
Rust, G. and Bide, M. (2000) The <indecs> Metadata Framework: principles, model and
data dictionary June 2000 WP1a-006-2.0,
http://www.doi.org/topics/indecs/indecs_framework_2000.pdf [accessed 2
October 2017]
Rutherford Appleton Laboratory (2015) The Alvey Programme,
www.chilton-computing.org.uk/inf/alvey/overview.htm [accessed 24 March 2017].
Salton, G., Allan, J. and Buckley, C. (1993) Approaches to Passage Retrieval in Full
Text Information Systems. In Proceedings of the 16th Annual International ACM
SIGIR Conference on Research and Development in Information Retrieval, SIGIR ’93,
New York, NY, ACM, 49–58.
Salton, G. and Yang, C. S. (1973) On the Specification of Term Values in Automatic
Indexing, Journal of Documentation, 29 (4), 351–72.
Schneier, B. (2015) Data and Goliath: the hidden battles to collect your data and control
your world, New York, NY, W. W. Norton.
Schreiber, G. and Raimond, Y. (2014) RDF 1.1 Primer, W3C,
www.w3.org/TR/rdf11-primer [accessed 26 April 2017].
Shannon, C. E. (1948). A Mathematical Theory of Communication, Bell System
Technical Journal, 27 (3), 379–423.
Shannon, C. E. and Weaver, W. (1949) The Mathematical Theory of Communication,
Urbana, IL, University of Illinois Press.
Sheriff, A. M., Bouchlaghem, D., El-Hamalawi, A. and Yeomans, S. (2011)
Developing a Metadata Standard for Multimedia Content Management: a case
study, Architectural Engineering and Design Management, 7, 157–76.
Singh, J. and Kumar, V. (2014) Virtual Appliances-Based Framework for Regulatory
Compliances in Cloud Data Centers, IUP Journal of Information Technology, 10 (1),
30–47.
Singhal, A. (2012) Introducing the Knowledge Graph: things, not strings, Google Official
Blog, https://googleblog.blogspot.co.uk/2012/05/introducing-knowledge-graph-
things-not.html [accessed 4 August 2017].
Skinner, G., Han, S. and Chang, E. (2005) A New Conceptual Framework within
Information Privacy: meta privacy. In Hao, Y., Liu, J., Wang, Y., Cheung, Y.-M.,
Yin, H., Jiao, L., Ma, J. and Jiao, Y-C. (eds) Computational Intelligence and Security,
Pt 2, Lecture Notes in Artificial Intelligence, Springer, 55–61.
SLA (2017) SLA Taxonomy Division, http://taxonomy.sla.org/ [accessed 2 February
2017].
Smiraglia, R. (2005) Introducing Metadata. In Smiraglia, R. (ed.) Metadata: a
cataloger’s primer, Binghampton, NY, Haworth Information Press, 1–15.
Solove, D. J. (2011) Nothing to Hide: the false tradeoff between privacy and security, New
Haven, CT, Yale University Press.

Sparck-Jones, K. (1972) A Statistical Interpretation of Term Specificity and its
Application in Retrieval, Journal of Documentation, 28 (1), 11–21.
Sperberg-McQueen, C. M. and Thompson, H. (2014) XML Schema,
www.w3.org/XML/Schema [accessed 9 June 2015].
Sponsors of Schema.org (2017) Schema.org, http://schema.org [accessed 19 April
2017].
Stock, W. G. and Stock, M. (2013) Handbook of Information Science, Berlin, Boston, De
Gruyter Saur.
Sun, S., Lannom, L. and Boesch, B. (2003) Handle System Overview. RFC 3650,
www.ietf.org/rfc/rfc3650.txt [accessed 19 August 2015].
Sundgren, B. (1973) An Infological Approach to Data Bases. PhD Thesis. University of
Stockholm.
Sweet, L. E. and Moulaison, H. L. (2013) Electronic Health Records Data and
Metadata: challenges for big data in the United States, Big Data, 1 (4), 245–51.
Syn, S. Y. and Spring, M. B. (2013) Finding Subject Terms for Classificatory Metadata
from User-generated Social Tags, Journal of the American Society for Information
Science & Technology, 64 (5), 964–80.
Tallon, P. P., Ramirez, R. and Short, J. E. (2013) The Information Artifact in IT
Governance: toward a theory of information governance, Journal of Management
Information Systems, 30 (3), 141–78.
Taxonomy Warehouse (2017) Taxonomy Warehouse, www.taxonomywarehouse.com
[accessed 2 February 2017].
TEI Consortium (2016) TEI P5: Guidelines for Electronic Text Encoding and Interchange,
Version 3.1.0, www.tei-c.org/Guidelines/P5 [accessed 31 May 2017].
UK Parliament (2016) Investigatory Powers Act.
United States (1791), US Constitution Amendment IV.
University of Essex (2017) UK Data Archive, www.data-archive.ac.uk [accessed 1
March 2017].
University of Nottingham (2014) Directory of Open Access Repositories – OpenDOAR,
www.opendoar.org/index.html [accessed 23 February 2017].
University of Toronto Library (2015) Subject Access Collection,
http://current.ischool.utoronto.ca/collections/special#SAS [accessed 2 February
2017].
US Environmental Protection Agency (2017) System of Registries,
https://ofmpub.epa.gov/sor_internet/registry/sysofreg/home/overview/home.do
US National Institutes of Health (2017) ClinicalTrials.gov Protocol Registration Data
Element Definitions for Interventional and Observational Studies,
https://prsinfo.clinicaltrials.gov/definitions.html [accessed 2 March 2017].
van Dijck, J. (2013) The Culture of Connectivity: a critical history of social media, Oxford,
REFERENCES 255
Oxford University Press.

van Hooland, S. and Verborgh, R. (2014) Linked Data for Libraries, Archives and
Museums: how to clean, link and publish your metadata, London, Facet Publishing.
van Rijsbergen, C. J. (1979) Information Retrieval, 2nd edn, London, Butterworth.
Vellucci, S.L. (1998) Metadata. In M. E. Williams, ed. Annual Review of Information
Science and Technology, Medford, NJ, Information Today Inc., 33, 187–222.
Visual Resources Association (2015) VRA Schemas and Documentation,
www.loc.gov/standards/vracore/schemas.html [accessed 5 May 2017].
W3C (2012) OWL 2 Web Ontology Language Primer, 2nd edn, W3C Recommendation,
www.w3.org/TR/2012/REC-owl2-primer-20121211 [accessed 2 February 2017].
W3C (2013) PROV-O: The PROV Ontology, www.w3.org/TR/2013/REC-prov-o-
20130430/ [accessed 4 October 2017].
W3C (2014a) Data Catalog Vocabulary (DCAT), www.w3.org/TR/vocab-dcat/
[accessed 3 December 2017].
W3C (2014b) HTML5, www.w3.org/TR/html5 [accessed 19 April 2017].
W3C (2015) SKOS Datasets, www.w3.org/2001/sw/wiki/SKOS/Datasets [accessed 9
February 2017].
W3C (2016) Extensible Markup Language (XML), www.w3.org/XML [accessed 19
April 2017].
Walcott, D. (1990) Omeros, 1st edn, London, Faber & Faber.
Weber, M. B. and Austin, F. A. (2011) Describing Electronic, Digital, and Other Media
Using AACR2 and RDA: a how-to-do-it manual and CD-ROM for librarians, London,
Facet Publishing.
Weibel, S. L. (1998) The Metadata Landscape: conventions for semantics, syntax and
structure in the internet commons. In Metadiversity, held in Natural Bridge, VT,
November 1998. Philadelphia, PA, National Federation of Abstracting and
Information Services.
Welsh, A. and Batley, S. (2012) Practical Cataloguing: AACR, RDA and MARC 21,
Whalen, M. (2016) Rights Metadata Made Simple. In Baca, M. (ed.) Introduction to
Metadata 3.0, Los Angeles, CA, Getty Research Institute.
White, M. (2016) The Value of Taxonomies, Thesauri and Metadata in Enterprise
Search, Knowledge Organization, 43 (3), 184–92.
Wilkinson, M. D. et al. (2016) The FAIR Guiding Principles for Scientific Data
Management and Stewardship, Scientific Data, 3 (18)
https://www.nature.com/articles/sdata201618 [accessed 2 October 2017].
Will, L. and TaxoBank (2013) Software for Building and Editing Thesauri,
www.taxobank.org/content/thesauri-and-vocabulary-control-thesaurus-software
[accessed 2 February 2017].
Wilson, A. (2010) How Much Is Enough: metadata for preserving digital data,
Journal of Library Metadata, 10 (2–3), 205–17.
Winter, J. (2008) Exploiting XML Structure to Improve Information Retrieval in Peer-

to-peer Systems, In Proceedings of the 31st Annual International ACM SIGIR
Conference on Research and Development in Information Retrieval, SIGIR ’08, New
York, NY, ACM.
WIPO (2015) WIPO, www.wipo.int [accessed 13 October 2015].
WIPO (2017) WIPO-Administered Treaties, www.wipo.int/treaties/en [accessed 4
October 2017].
World Economic Forum (2011) Personal Data: the emergence of a new asset class,
Geneva.
Yahoo! Inc. (2017) Flickr, www.flickr.com [accessed 20 April 2017].
Zavalina, O. L. (2011) Contextual Metadata in Digital Aggregations: application of
collection-level subject metadata and its role in user interactions and information
retrieval, Journal of Library Metadata, 11, 104–28.
Zeng, M. L. and Qin, J. (2008) Metadata, 1st edn, New York, NY and London, Neal-
Schuman Publishers.
Zeng, M. L. and Qin, J. (2015) Metadata, 2nd edn, Chicago, IL, ALA Neal-Schuman.
Index
AACR2 5, 8, 40, 42, 56 provenance 43, 135–6, 234

ABC Ontology 42–4, 48 standards for description see EAD
A-Core model 180–1 and ISAD(G)
acquisition of library resources ARK (Archival Resource Key) 84–5
117–18 audiovisual materials 83–4
administrative metadata 14, 69, 71, retrieval 108–11, 141, 233
109, 180–1 standards 65
AIP (Archival Information Package) (see also images)
47, 120 audit trail
Alexandria Library 4, 227 e-discovery 156
Amazon 145, 148, 206 governance 17, 122, 235, 236
analysis of requirements 166–7 provenance 135, 136, 137–8, 152–3
Anglo American Cataloguing Rules authentication of users 14, 116, 122
see AACR2 authenticity of data 152–3, 156, 236
application profiles 174, 230 author names 90, 191 (see also
DCAP 54 Creator data element)
Dublin Core 149 authority lists 90–1, 191–2, 193
ONIX 132, 168 automatic indexing 107, 188, 224,
Singapore Framework 54, 170, 183 229
Archival Information Package see automatic metadata generation 169,
AIP 193
Archival Resource Key see ARK
archives barcodes 80, 84, 86 (see also EAN
data archives see research data codes)
collections Bayesian analysis 98, 99, 102
politics and ethics 228, 231 BBC Domesday Book Project 119–20
BIBFRAME 8, 57–8, 124 controlled vocabularies 27, 193, 194

and crosswalks 175, 178, 179 BIBFRAME 58
bibliographic citation 88, 91 DCAT (Data Catalog vocabulary)
bibliographic data 56–7, 145–6 178, 209
big data 203–6, 219, 227, 237 Dublin Core 52, 53, 54
(see also open data repositories encoding and maintenance 164
and research data collections) information retrieval 104
book provenance 136–7 IPTC PhotoMetadata Standard
book trade see publishing and the 69
book trade Schema.org 196–7
Boolean logic 99–100 taxonomies 185–6, 187
thesauri 188, 189, 190–1
cataloguing 118 cookies 140, 142–3, 225
history 4–7 costs and functionality,
library catalogues 4, 6, 29–30, 122–3 interoperability 174
union catalogues 6, 31 Creative Commons 53, 128, 234
cataloguing rules 31, 88, 94, 164 images 146
AACR2 8, 40, 42, 56 music industry 148
authority lists 191 rights management 133–4
history 5 Creator data element 51, 89–91, 93
RDA 31, 40, 90, 91 crosswalks 164, 175–9, 183
CDWA Categories for the
Description of Works of Art) data archive see research data
177–8 collections
classification 179, 190, 194, 224 databases of metadata 19, 26–7
information retrieval 106, 186 Data Catalog vocabulary see DCAT
library catalogues 29 data elements, Dublin Core 24–5,
standards 53 52–4
cloud services 152, 158, 206 data mining 205, 206
complex objects 70–4 data models 35
compliance 17, 151, 152, 153, 154–6 ABC Ontology 42–4
data security 156-8 DCMI Resource Model 39–40
e-discovery 156 indecs 44–5
freedom of information 154 LRM (Library Reference Model)
privacy and data protection 154-6 40–2, 55, 89
sectoral compliance 158–9 OAIS (Open Archival Information
computational models of retrieval System) 46–7, 120, 125
97, 107 RDF (Resource Description
Content-Based Image Retrieval Framework) 36–9
(CBIR) 108, 110 (see also images UML (Unified Modelling
– retrieval) Language) 36
INDEX 259
data portals 106 (see also open data document mark-up 19–22 (see also
repositories and portals) mark-up languages)
data protection see privacy and data DOIs (Digital Object Identifiers) 44,
protection 79, 81–3, 206, 215
data retrieval 95 Domesday Project see BBC
Date data element 26, 52, 59, 91–2, 93 Domesday Book Project
DCAP (Dublin Core Application DTDs (Document Type Definitions)
Profile) 54 23–4, 62
DCAT (Data Catalog vocabulary) Dublin Core 7, 39–40, 65, 71–2
178, 207, 209 (see also application profiles 54, 149, 170
application profiles) crosswalks 175–6, 178
de facto standards 50, 148, 168, 233 data elements 24–5, 52–4
Description data element 92–3, 94 DCMI 51–4
descriptive metadata 14, 69–70, 71, rights management 128–9
86–7, 88–93, 94 Dublin Core Application Profile see
DIDL (Digital Items Declaration DCAP
Language) 132–3 Dublin Core Metadata Initiative see
Digital Curation Centre 167 Dublin Core – DCMI
lifecycle model 116, 118, 165
Digital Items Declaration Language EAD (Encoded Archival
see DIDL Description) 62, 74, 115
Digital Object Identifiers see DOIs EAN codes (European Article
digital objects 19, 67, 70-1, 82, 133, 218 Number) 80, 81, 84 (see also
METSRights 129 barcodes)
PREMIS 120, 121, 129 EBUCore 65
provenance 136, 137, 138 ECM (Enterprise Content
Schema.org 196 Management) systems 26–7
(see also electronic documents) e-commerce 16, 141, 235
digital resources 113, 128, 137, 194 electronic transactions 139–40,
preservation 118–19 211–12
DIP (Dissemination Information images 146–7
Package) 47 indecs 44–5, 143–4
Directory of Open Access music industry 147–8
Repositories see OpenDOAR ONIX 118, 143–4, 145–6
disposal of library materials 118, publishing and the book trade 48,
123–4 144–8
records 16, 28, 29, 115 e-discovery 156, 159
Dissemination Information Package EDRM (Electronic Document and
see DIP Records Management) 16, 27,
Document Type Definitions see 29, 186
DTDs creating metadata 193
e-GMS (e-Government Metadata FRBR (Functional Requirements for

Standard) 149, 230 Bibliographic Records) 78–9, 87,
e-government 139–40, 148–9, 235 89, 136
e-Government Metadata Standard freedom of information 15, 17, 151,
see e-GMS 154 (see also compliance)
Electronic Document and Records friend of a friend see FOAF
Management see EDRM Functional Requirements for
electronic documents 16, 19, 26 Authority Data see FRAD
authentication 152, 156, 236 Functional Requirements for
provenance 134, 135 Bibliographic Records see FRBR
(see also digital objects) funding 216, 229
electronic transactions 139–40, 211–12 fuzzy searching 98, 99, 100
embedded metadata 8, 11, 26, 99,
133, 194 General International Standard for
Encoded Archival Description see Archival Description see
EAD ISAD(G)
encoding schemes geographical information systems
authority lists 191–4 6–7
controlled vocabularies 186–8, geographic coverage 52, 149, 227
190–1 governance 17, 151–9, 235–6
folksonomies 199–201
ontologies 194–8 hashtags see Twitter hashtags
taxonomies 185–6, 188 HIDDEL (Health Information
thesauri 188–90 Disclosure, Description and
Enterprise Content Management Evaluation Language) 157–8
systems see ECM systems HTML (Hypertext Mark-up
ethics 158, 221–6 Language) 22, 99, 196
European Article Number see EAN
European Union Open Data Portal identifiers see resource identification
207–8 and description
EXIF 66, 109 IDF (Inverse Document Frequency)
Extensible Mark-up Language see function 101, 102
XML IEEE LOM (Learning Object
Metadata) 72–4, 231
federated search service 217, 218 IIIF (International Image
file plans 28, 123, 156, 186 (see also Interoperability Framework)
records management) 67–9
FOAF (friend of a friend) 62–3, 195 iLumina project 174–5
Format data element 92, 93 images
FRAD (Functional Requirements for Creative Commons 146
Authority Data) 40, 55, 191 e-commerce 146–7
INDEX 261
repositories 33 information silos 205, 206, 209, 217,

retrieval 104, 108–11, 141, 233 219
standards 64–70 silo-based searching 218
(see also audiovisual materials) Information Theory see Shannon’s
importing metadata 164, 166, 169, Information Theory
174 institutional repositories 71–2,
indecs 35, 127, 130, 137, 145 217–19
e-commerce 44–5, 143–4 Intellectual Property Management
interoperability 171 and Protection see IPMP
rights management 130, 131, 132 intellectual property rights 89,
(see also ONIX) 127–38, 234–5
indexed-based discovery system International Image Interoperability
217, 219 Framework see IIIF
indexing 4, 16, 168 International Standard Audiovisual
authority lists 191, 193 Number see ISAN
automatic indexing 107, 188, 224, International Standard Bibliographic
229 Description see ISBD
image retrieval 109–10, 111, 233 International Standard Book Number
latent semantic indexing 98, 99, see ISBN
101, 193 International Standard Serial
pre-coordinate and post- Number see ISSN
coordinate indexes 190–1 International Standard Text Code
records management 114 see ISTC
retrieval 104 internet retrieval 104–6
self-indexing 199 interoperability 16, 43, 44, 170–3
Shannon’s Information Theory content standards 49–50
97-8, 205 costs and functionality 174
subject indexing 95, 106–7 institutional repositories 217
thesauri 189 normalising data 174–5
information inequality 223–4 research data collections 215
information lifecycle 113–17, 118, intranets 106–7
124, 125 (see also lifecycle) Inverse Document Frequency see
information retrieval IDF Inverse Document
Boolean logic 99–100 Frequency function
computational models of retrieval IPMP (Intellectual Property Manage-
97, 107 ment and Protection) 132
images 104, 108–11, 141, 233 IPTC Photo Metadata Standard
internet 104–5 69–70, 147
precision and recall 102–4, 190 ISAD(G) (General International
search engines and ranking Standard for Archival
105–6 Description) 60–2, 74, 124, 191
ISAN (International Standard resources 233–4

Audiovisual Number) 83 create or ingest 117–18
ISBD (International Standard distribute and use 122–3
Bibliographic Description) 5, 88, information lifecycle 113–17
172 preserve and store 118–19
ISBN (International Standard Book review and dispose 123–4
Number) 78, 79, 80–1, 85, 232 transform 124
ISMN (International Standard management of metadata 163, 236–7
Music Number) 84 application profiles 170
ISSN (International Standard Serial interoperability of metadata
Number) 83, 213 171–9
ISTC (International Standard Text metadata lifecycle 164–5
Code) 84, 85, 93 metadata security 181–2
project approach 165–9
JPEG2000 66 quality considerations 179–81
MARC (Machine Readable
KBART (Knowledge Bases and Cataloguing) 5–7, 71, 230
Related Tools) 59–60 MARC 21 55–7, 230, 233
knowledge management 323 interoperability 171, 173
latent semantic indexing 98, 99, 101, mark-up languages
193 HTML 22, 48, 88, 99, 110, 196
LaTeX 22 LaTeX 22
Learning Object Metadata see IEEE SGML 19, 20–2
LOM TEI 20, 22
library catalogues 4, 6, 29–30, 122–3 XML 20, 22–4, 38
library management systems 8, 118, Massive Open Online Courses see
124, 125 MOOCs
Library Reference Model see LRM metadata
lifecycle definition 9–10
Digital Curation Centre model embedded 11, 12, 26
116, 118, 165 examples of 11, 27–9, 32–5
information 113–17, 118, 124, 125 purposes of 11–17, 75–159, 232–6
project 165–6 registries 164, 179
workflow and metadata 164–5 security 181–2
linked data 206–9, 228 metadata elements
LRM (Library Reference Model) administrative metadata 180–1
40–2, 55, 89 characteristics 87–8
descriptive metadata 88–93
Machine Readable Cataloguing see Dublin Core 52–4
MARC rights management 128–9
management of information XML schemas 24–6
INDEX 263
Metadata Encoding and Transmission Standards) 64,

Transmission Standards see 70–1, 74, 127
METS METSrights 128, 129, 137
Metadata for Images in XML see MIX (Metadata for Images in XML)
MIX 65
metadata harvesting 200, 219 MODS (Metadata Object
interoperability 172 Description Schema) 56, 58–9,
OAI-PMH 71–2, 106, 217–18 70, 71
Metadata Object Description crosswalks 175, 176
Schema see MODS MOOCs (Massive Open Online
metadata registries 164, 179 Courses) 73–4
metadata schemas see metadata Moving Pictures Expert Group see
standards MPEG
metadata security 181–2 MPEG (Moving Pictures Expert
metadata standards 49-74 Group) 65, 130, 211
BIBFRAME 8, 57-8, 124 MPEG-7 65, 109
Dublin Core 7, 39-40, 65, 71-2 MPEG-21 127, 130, 137
EAD 62, 74, 115 rights management 132–3
EBUCore 65 music industry 147–8
e-GMS 149, 230 Creative Commons 148
EXIF 66, 109 rights management 148, 234–5
FOAF 62-3
IEEE LOM 72-4, 231 name authorities 79, 94, 191 (see also
IIIF 67-9 authority lists)
indecs 35, 127, 130, 137, 145 namespace 24, 25, 26, 36, 38, 52
IPTC 69-70, 147 non-textual materials 64–70, 109,
ISAD(G) 60-2, 74, 124, 191 132–3
MARC 21 55-7, 230, 233 normalising data, interoperability
METS 60, 70-1, 74, 127 174–5
MIX 65
MODS 56, 58-9, 70, 71 OAI-PMH (Open Archives Initiative
ONIX 51, 139 – Protocol for Metadata
OpenGraph 63 Harvesting) 71–2, 106, 217–18
PBCore 65, 74 OAIS (Open Archival Information
PREMIS 120-1, 125, 137, 234 System) 46–7, 120, 125
RDA 8, 74, 191 ODRL (Open Digital Rights
VRA Core 64-5, 74 Language) 129–31, 137
(see also standards) ONIX 51, 139
METeOR Metadata Online Registry bibliographic data exchange 145–6
179 e-commerce 16, 45, 118, 144–5
METS (Metadata Encoding and and indecs 143–4
rights management 131–2 intellectual property rights 129

online behavioural advertising 141–3, metadata element 178
211–12, 235, 236 (see also preservation 13, 14, 15, 46, 47
e-commerce and social media) digital objects 118–19
ontologies 48, 106, 194, 208, 236 information lifecycle 113
ABC ontology 42–4 OAIS for preservation 120
FOAF (Friend of a Friend) 62–3 PREMIS for preservation 120–1
OWL (Web Ontology Language) preserve and store 118–21
169, 195–6 privacy and data protection 122,
SKOS (Simple Knowledge 154–6, 182
Organization System) 195–6 ethics 221-2
open access 213, 215, 234 meta privacy 158, 182
Open Archival Information System records management 156
see OAIS social media 156
Open Archives Initiative – Protocol probabilistic model 98, 102, 105, 210
for Metadata Harvesting see project lifecycle 165–6
OAI-PMH Project Open Data see POD
open data repositories 206–9 PROV 135, 137, 138
POD (Project Open Data) 209 provenance 46, 87, 120, 134, 181
Open Digital Rights Language see books and printed material 136–7
ODRL digital objects 136, 137, 138
Open Discovery Initiative 212–13 metadata 137
OpenDOAR (Directory of Open PROV 135, 137, 138
Access Repositories) 32–3, 216, records and archives 43, 135–6,
217 234
Open Graph Protocol 63, 74, 211, 231 publishing and the book trade 48,
OWL (Web Ontology Language) 144–8, 230, 235
195-6 indecs 44–5, 143–4
ONIX 118, 143–4, 131–2
PBCore 65, 74 rights management 127–8
PDI (Preservation Description purposes of metadata 11–17, 75–159,
Information) 46–7, 120 232–6
PhotoMetadata Project 109
POD (Project Open Data) 209 quality management 165, 179–80,
portals 106, 206–9 181
power and ownership of metadata
space 226–9 radio-frequency identifiers see RFIDs
precision 102–4, 190 RDA (Resource Description and
pre-coordinate and post-coordinate Access) 8, 74, 191
indexes 190–1 LRM (Library Reference Model)
PREMIS 120–1, 125, 137, 234 40–2
INDEX 265
metadata elements 87, 88, 89, 90, ONIX 131–2

91, 94 publishing and the book trade
standards 54–5 127–8
RDF (Resource Description
Framework) 36–9, 58, 63, 170 Schema.org 196–7
open data repositories 206, 207 schemas 127, 128, 129, 130
recall 102–4, 190 crosswalks 175–9
records management 27–9 selecting 164, 166, 167–8
authentication 152 XML 24–6
file plans 28, 123, 156, 186 SCORM (Sharable Content Object
information lifecycle 114–15 Reference Model) 73
metadata elements 117 search engines 97, 98, 99, 100, 228
privacy and data protection 156 data elements 107
provenance 134, 135–6 optimisation 140–1
review and dispose 123–4 ranking 105–6
REGNET 158, 159 Schema.org 196–7
regulatory compliance see sectoral compliance 158-9 (see also
compliance compliance)
requirements for metadata 166–7 self-indexing 199
semantic web 205, 228, 229
research data collections 212–19 ABC ontology 42
Resource Description and Access see linked data 206-9, 228
RDA OWL 195-6
Resource Description Framework RDF 8, 74, 191
see RDF Schema.org 196
resource discovery 12–13, 15, 86 SKOS 178, 195-6
resource identification and serials 55, 83, 145
description 53, 77–84, 88–93, SGML (Standard Generalized
232 Mark-up Language) 19, 20–2
retrieval see information retrieval Shannon’s Information Theory 97–8,
RFIDs (radio-frequency identifiers) 111, 205
85–6, 232 Sharable Content Object Reference
RFID (Radio-Frequency IDentity) Model see SCORM
85–6, 232 silo-based searching 218
rights management 69–70 Singapore Framework 54, 170, 183
Creative Commons 133–4 SIP (Submission Information
Dublin Core 128–9 Package) 14, 47
indecs 130, 131, 132 social media 140, 222, 234
metadata elements 128–9 metadata standards 62–4
MPEG-21 132–3 online behavioural advertising
music industry 148, 234–5 104, 156, 211–12, 235
social tagging and folksonomies TEI 20, 22

199–201 thesauri 186–91, 193
Standard Generalized Mark-up Title data element 88–9, 93
Language see SGML Turtle (Terse RDF Triple Language)
Standards 49–51 48
complex objects 70–5 Twitter hashtags 64, 226 (see also
data portals 208–9 social media)
de facto 50, 148, 168, 233
e-government 149 UML (Unified Modelling Language)
library and Information 54–62 36, 39
non-textual materials 64–70, 109 Uniform Resource Identifier see URI
preservation 51, 129, 135, 143–6, union catalogues 6, 31
148 Universal Resource Locators see
social media 62–4, 195 URLs
(see also metadata standards) Universally Unique Identifier see
standards development 50, 230, 231 UUID
stylesheets (in SGML) 21, 23 URI (Uniform Resource Identifier)
DTD 23, 24 37, 38–9, 79
subject indexing 95, 106–7 URLs (Universal Resource Locators)
Submission Information Package see 79, 82–3, 204
SIP user education 225–6
surrogacy 119 UUID (Universally Unique
synonyms 105, 186, 191, 193, 200 Identifier) 80
synonym rings 188–90
vector spaces for ranking 99, 101
tags and tagging 11, 21, 24, 38, 56–7, VLEs (Virtual Learning
59 Environments) 71, 72, 87
EAD schema 62 VRA Core 64–5, 74
e-commerce 140
image tags 141 Web Ontology Language see OWL
music industry 147, 148 (Web Ontology Language)
open data repositories 209 Wikileaks 3, 4, 154
RFID tags 85 WIPO (World Intellectual Property
Schema.org 196, 197 Organization) 127–8, 132
search engine optimisation 140, workflow and metadata lifecycle
141 164–5
social tagging and folksonomies WorldCat union catalogue 31, 32,
104, 110, 169, 172, 174, 199–201 217
Twitter hastags 64, 226 World Intellectual Property
taxonomies and encoding schemes Organization see WIPO
185–201
INDEX 267
XML (Extensible Mark-up Images in XML); XSDL)

Language) 20, 22–4, 38 XSDL (XML Schema Definition
schemas 24–6 Language) 24
(see also MIX (Metadata for

Handbook

Uploaded by

Copyright:

Available Formats

Handbook

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Handbook

Uploaded by

Copyright:

Available Formats

Haynes 4th proof 13 December 2017 13/12/2017 15:37 Page i

Every purchase of a Facet book helps to fund CILIP’s advocacy,

© David Haynes 2004, 2018

Published by Facet Publishing

Facet Publishing is wholly owned by CILIP: the Library and Information

British Library Cataloguing in Publication Data

ISBN 978-1-85604-824-8 (paperback)

First published 2004

Text printed on FSC accredited material.

List of figures and tables ix

PART I METADATA CONCEPTS 1

2 Defining, describing and expressing metadata 19

VI METADATA FOR INFORMATION MANAGEMENT AND RETRIEVAL

OAIS – Online exchange of data 46

PART II PURPOSES OF METADATA 75

6 Retrieving information (Purpose 2) 95

7 Managing information resources (Purpose 3) 113

8 Managing intellectual property rights (Purpose 4) 127

9 Supporting e-commerce and e-government (Purpose 5) 139

10 Information governance (Purpose 6) 151

PART III MANAGING METADATA 161

12 Taxonomies and encoding schemes 185

13 Very large data collections 203

VIII METADATA FOR INFORMATION MANAGEMENT AND RETRIEVAL

What is big data? 205

14 Politics and ethics of metadata 221

List of figures and tables

X METADATA FOR INFORMATION MANAGEMENT AND RETRIEVAL

6.1 Resolution power of keywords 96

XII METADATA FOR INFORMATION MANAGEMENT AND RETRIEVAL

David Haynes, August 2017

XIV METADATA FOR INFORMATION MANAGEMENT AND RETRIEVAL

4 PART I METADATA CONCEPTS

Fundamental principles of metadata

In Western Europe library cataloguing developed in the ecclesiastical and,

6 PART I METADATA CONCEPTS

exchanging cataloguing data, which helped to reduce the cost of cataloguing

• objects (real-world phenomena)

8 PART I METADATA CONCEPTS

Looking at existing trends, therefore, metadata is becoming more widely

The politics of information, and in particular metadata, have become more

10 PART I METADATA CONCEPTS

metadata for controlling access; or content ratings metadata, a key component of

In his introduction to Metadata: a cataloger’s primer Richard Smiraglia provides

Metadata are structure, encoded data that describe the characteristics of

A further description is proposed to cover the range of situations in which

What does metadata look like?

Title: Sapiens: a brief history of humankind / Yuval Noah Harari.

12 PART I METADATA CONCEPTS

Figure 1.1 Metadata from the Library of Congress home page

Metadata Initiative. He recognises that metadata is used for administering

Table 1.1 Day’s model of metadata purposes

14 PART I METADATA CONCEPTS

Category Deﬁnition Example

Administrative Metadata used in managing • Acquisition and appraisal information

Gilliland model refers to Technical metadata ‘related to how a system

The six-point model

1 Resource identification and description – This is particularly important in

16 PART I METADATA CONCEPTS