Applying Chatbots to the Internet of Things:

Opportunities and Architectural Elements

Abstract—Internet of Things (IoT) is emerging as a significant Billion [2, 3]. Among them more than half of all IoT endpoints
technology in shaping the future by connecting physical devices or in the consumer space alone. Hence IoT is a phenomenon which
things with internet. It also presents various opportunities for is certain to play a major role in our daily interaction with the
intersection of other technological trends which can allow it to digitally connected world.
become even more intelligent and efficient. In this paper we focus
our attention on the integration of Intelligent Conversational
Software Agents or Chatbots with IoT. Literature surveys have
A. Scope of Internet of Things
looked into various applications, features, underlying technologies
and known challenges of IoT. On the other hand, Chatbots are The literature presents various ways to define the Internet of
being adopted in greater numbers due to major strides in Things. The RFID group defines Internet of Things as “world-
development of platforms and frameworks. The novelty of this wide network of interconnected objects uniquely addressable,
paper lies in the specific integration of Chatbots in the IoT based on standard communication protocols”. ITU [4] defines it
scenario. We analyzed the shortcomings of existing IoT systems as “a global infrastructure for the information society, enabling
and put forward ways to tackle them by incorporating chatbots. A advanced services by interconnecting (physical and virtual)
general architecture is proposed for implementing such a system, things based on existing and evolving interoperable information
as well as platforms and frameworks – both commercial and open and communication technologies”.
source – which allow for implementation of such systems.
Identification of the newer challenges and possible future While considering the broad vision of IoT, this paper focuses
directions with this new integration, have also been addressed. on the perspective of connected things and applications for those
things. To do this we simply create a separation of concern
Keywords—Internet of Things, Chatbots, Human- between the fragmented lower Open System Interconnection
Computer Interaction, Conversational User Interfaces, (OSI) layers of IoT and the unifying adopted upper layers of IoT
Software Agents communication which uses the World Wide Web and its
standard network protocols.
I. INTRODUCTION
The entire IoT system consists of Sensors (temperature,
The Internet of Things (IoT) is not just a well-recognized light, motion, etc.), Actuators (displays, sound, motors, etc.),
phenomenon but one that is shaping the digital age. It Computation (programs and logic), and Communication
introduces an era of interconnected smart objects or ‘things’ interfaces (wired or wireless). However, based on established
developed upon existing internet architectures. By using unique advantages presented in the literature [5, 6, 7, 8], our scope will
addressing schemes and standard communication protocols, be limited to interaction with IoT through Web Application
IoT interconnects these things or objects thereby creating a Programming Interfaces (API) and in particular Hypertext
varied range of technologies are able to interact with each other Transfer Protocol (HTTP) based Representational State Transfer
and reach common goals [1]. (REST) Architectures. A popular approach of Web of Things
An essential goal of connecting various sensors, has been illustrated in Fig 1 based on [6].
actuators and services and collecting/processing data from them The Evans Data Corporation (EDC) Report: Internet of
is to generate situational awareness and enable machines and Things - Vertical Research Service study [9] reveals that more
human users to make sense of themselves and their surrounding than half of IoT developers connect to devices primarily through
environments. the cloud. The massive growth and acceptance of these cloud
The proliferation of IoT can be seen through adoption of based platforms such as IBM IoT Platform, AWS IoT, Microsoft
these “smart devices” in our daily life which include Azure IoT and Cisco IoT show that the new generation of IoT
applications in Manufacturing, Agriculture, Medical and applications concentrate on cloud based platforms with the
Healthcare, Transportation, Building and Home Automation and lower layers (Transfer, Transport and Network). Hence, this
Energy Management among others. A report by Gartner paper also proposes the use of IoT cloud based platforms in our
estimates that there will be over 20 Billion connected things in architectural design. This is discussed further in Section 4.
activity by 2020 with Cisco estimating the number to be over 50
 of these paradigms have largely been separate endeavors. We
discuss how using chatbots as intelligent conversational
interfaces can be used to address critical problems in IoT. We
also propose a high level conceptual architecture and discuss key
elements involved in communicating with an IoT system
through Chatbots. To explain in the context of real world
applicability, we put forth existing solutions to each of the
components in the architecture including frameworks, platforms
and specify open-source tools which can be used to build such a
system.
The rest of this paper is organized as follows: In section II
we discuss our motivation for introducing this novel concept of
Chatbots in Internet of Things and discuss other literature work
that has helped shape this concept. In Section III, we evaluate
Fig. 1. Web of Things as shown in [6] and examine the shortcomings and challenges of current IoT
systems and the opportunity for chatbots to address them.
B. Scope of Chatbots Section IV proposes a system design and the key architectural
This paper proposes the use of Intelligent Conversational elements. Finally, we present our concluding remarks by
Agents. We refer to these as simply Chatbots (also known as assessing opportunities and scope for future research and
Chatterbots or bots in general). Interestingly, there are many development in Section V.
definitions for Chatbots in close relation with Software Agents
(SA), Virtual Agents (VA) or Intelligent Personal Assistants II. MOTIVATION AND BACKGROUND STUDY
(IPA) in literature and these have often been used in conjunction The key to the massive adoption and diffusion of IoT
with each other. The term “Agents” itself has many definitions is the proliferation of Internet in our daily lives. We use the
but among the earliest and most well-known uses of the term is internet to search for information, check emails, consume
[10] - "A self-contained, interactive and concurrently-executing media, and connect with people via social networks and so
object, possessing internal state and communication capability."
much more. With around 40% of the global population (3.4
The scope of Software Agents can be most closely Billion) currently using the world wide web, this number is
associated with Chatbots and has been well documented in estimated to increase to 7.6 billion global internet users in 2020,
literature [11]. The following key properties have been a majority of which use mobile devices (phones, tablets,
associated with Software Agents [12]: (1) reactive, (2) pro- wearables etc.) [14]. Hence the internet has played a vital role
active and goal-oriented, (3) deliberative (4) continual (5) as a global backbone for information sharing, interconnection
adaptive (6) communicative, and (7) mobile. The purpose of this of physical objects with computing/networking capabilities for
paper is not to explore the various types of Software Agents and applications and services spanning numerous use cases.
agent based systems or its properties but rather propose the Internet alone, however, cannot address all issues of IoT. First,
solution to challenges faced in IoT through the use of the we will briefly discuss the challenges in IoT, and then mention
umbrella term for these Intelligent Conversational Agents,
the motivation for choosing intelligent conversational
Software Agents or Chatbots as we refer to them. It is also
interfaces.
important to note that Software Agent distinguishes itself from
Intelligent Agents (also known as Rational Agents). Intelligent A. Challenges in IoT
agents are not only computer programs. They can also be
Despite the wide scale efforts to popularize IoT, it still
machines, humans or anything that is capable of a goal directed
offers many practical challenges. Primarily, IoT systems
behavior [13].
operate in isolated technology or vendor specific silos which
Typically Chatbots are classified into two types: (1) inhibits capability, value, and interoperability and create a
Chatbots that function based on Rules (2) Chatbots that function widely disparate area [15]. Specifically, by restricting
based on Artificial Intelligence. Chatbots that function on rules heterogeneous devices (home appliances, mobile phones,
are often limited as they are only as smart as they are embedded devices etc.), sensors and services to communicate
programmed. On the other hand, AI based Chatbots give the with each other across interconnected networks, possibilities of
impression of being “intelligent” as they are capable of countless applications are hindered.
understanding natural language, not just pre-defined commands Secondly, the sheer number of connected things has
but get smarter as they interact more due to their ability to already started to create problems in application, device and
maintain different states. Based on this, concepts such as Virtual
data management in IoT [16]. To address this issue, IoT
Agents and Intelligent Personal Assistants (IPA) have come up,
which uses natural language processing, as well as speech platforms (such as Cisco IoT, IBM IoT, Microsoft Azure IoT,
recognition techniques. For example, Apple Siri, Amazon AWS IoT) offer scalable, distributed cloud based services in
Alexa, Microsoft Cortana and Google Assistant. order to allow businesses to quickly connect to an established
infrastructure, service or software without having to worry
In this paper we present a novel paradigm combining these about the backend complexities. While IoT Cloud is a step in
two disparate concepts in a single solution. However, the studies the right direction, offering many advantages, it still presents
many challenges particularly in interoperability which has led
to the issues of platform fragmentation [17, 18].

IoT systems also face a challenge of unifying User

Interfaces (UI). It becomes increasingly difficult on users to
keep track and access multiple applications, dashboards for
every new “IoT object” in their ecosystem [19]. Hence unifying
experiences across multiple connected things and providing
them with a high degree of smartness for improved user
experience is a key challenge.
B. Relevance of Conversational User Interfaces
According to reports, Chat interfaces which are used in
Instant Messaging (IM) platforms (Such as Facebook
Messenger, Slack, Kik, and Telegram) have been immensely Fig. 2. Sample of a IoT Chatbot-User conversation
popular and continue to show steady growth. IM services have
III. OPPORTUNITIES FOR CHATBOTS IN IOT
more active users than any other internet application including
social networks, mailing applications [20]. This report shows The shortcomings of modern IoT systems can be broadly
how the top ten messaging platforms alone account for nearly classified into two types: (i) Technology Centric Challenges and
4 Billion users. The global acceptance of chat based interfaces (ii) Human Centric Challenges. With the help of sample chatbot-
allows for ease of adoption and diffusion of newer technologies user conversations given in Fig. 2, we discuss the opportunities
built on top of the pre-existing platforms (such as Chatbot for Chatbots and demonstrate ways in which Chatbots can
Applications). Therefore the global proliferation of chat overcome challenges in IoT.
platforms only furthers the motivation to develop interesting Use Case (A)
applications and use cases with chatbots. User: “Keep the living room temperature comfortable”
Chatbot: “The weather outside is a cool 17 degrees Celsius.
On a different note, advancements made in the areas of Setting temperature in the living room to 21.4 degree Celsius.”
Artificial Intelligence, especially Natural Language Processing Use Case (B)
User: “How much is my car charged”
have furthered the efficiency and quality of Chatbots in terms
Chatbot: “The Tesla Model S is currently 40% charged. 3 Hours 10
of its conversational simplicity, easy adaptability and minutes to full charge.”
capabilities in allowing the user to make complex requests Use Case (C)
through simple natural language. User: “Turn the light on in the guest bedroom”
Chatbot: “Which light would you like to have turned ON? The
However, the similarity of both Chatbots and IoT lies in Lamp or Table Light?”
adoption of their services through relatively simple, often User: “Both”
RESTful Web APIs. The rationale behind it has been based on Use Case (D)
the following observations: User: ‘Help me setup my new device’
Chatbot: “Here is some help to guide you through the setup”
(1) Developers can take an API or service-oriented
“Which device would you like to setup?
approach to development for both IoT as well as 1) Smart Lock 2) Smart Kettle 3) Smart light?”
Chatbots. This means that application development User: 1
methodologies would be the same with both embedded Chatbot: “Ok, Enter your secret passcode for the smart lock”
devices as with any web service (including Chatbots) User: “*****”
that use Web APIs and in particular using RESTful Chatbot: “Done. Smart Lock is now setup.”
architectures. Use Case (E)
(2) Chatbot applications just like IoT applications can be Chatbot: “The monitoring service indicates that the smart lock has
designed and deployed on cloud platforms which allow been offline for over 24 hours.”
Chatbot: “Would you like me to report the issue to the Smart Lock
simple development and deployment without concern
Vendor?”
about the underlying technologies such as Transport User: “No, I want to talk to a human”
layer, Storage, Processing etc. Human-Operator: “I can see the issue you are facing. I will try to
(3) Owing to HTTP RESTful standards and protocols, it resolve it remotely.”
becomes technologically feasible and simple to
integrate chatbot applications into IoT systems using A. Technology Centric Challenges of IoT
application layer as the only concerned medium 1) Data Management
A key challenge in the realm of IoT, is managing the
This ease of integration is a key motivation to develop vast amount of big data being generated, as IoT sensors are
platforms and frameworks which can synchronize chatbot becoming easily affordable. Not only is the data generated by
applications within IoT platforms and frameworks. the sensors large but also diverse (varying in quality and type)
and multimodal (temperature, light, sound, video, etc.) in
nature. While data deluge is one challenge, drawing insights 2) Device and Application Management
from the data and being able to present it in a timely, A key challenge of IoT has been the fragmentation of
understandable way is a much larger challenge. technology [17, 18]. Having application interoperability
between heterogeneous devices from a single remote (mobile
The situation can be best illustrated by the well-known device or operation terminal) is especially uncommon. For
Knowledge Hierarchy also called the DIKW (Data, Example, consider the situation where a smart light and a
Information, Knowledge, Wisdom) Pyramid in the context of Heating Ventilation and Air Conditioning (HVAC) system
IoT [21, 22] which calls for solutions to tackle the difficult belong to the same network and environment yet may have
challenges as one moves up in the pyramid. The data gets different user control terminals which are mutually independent
smaller but becomes more difficult to gain abstractions and entities, unaware of each other nor able to control or
perceptions (Knowledge), which is required to derive communicate with each other.
actionable intelligence (Wisdom). Chatbots are attempting to
solve the problems of data and information management by Chatbots are built on IM platforms (such as Facebook
mainly addressing the upper layers of the DIKW pyramid. Messenger and Slack) which support multiple different chatbot
a) Data Context applications. A single chatbot application as well can use
Processing and analyzing of IoT data can be solved through unique HTTP REST APIs pertaining to different IoT devices.
the many “big data” solutions and cloud platforms which offer Chatbots can thus act as a single interface for communication
storage and computing infrastructure to accomplish the task. between single purpose devices (eg. Controlling two smart
These existing IoT cloud solutions handle Data source and lights), heterogeneous devices (eg. Controlling a HVAC and a
transmission challenges. However, a major challenge of existing Smart Car) and even different IoT ecosystems (eg. Controlling
IoT systems is conveying data about the various interconnected Smart home devices and Smart Retail devices) in the case of
devices (sensors, objects etc.) back to the user in a simple cloud based IoT. For example, in the Use cases above, the same
humanly understandable way. This requires context, which is chatbot is being used to converse with multiple heterogeneous
achieved by enabling Chatbots to understand the true intent of devices. Provided the right permissions are available it can
the user query and collect and process information from their communicate with Public IoT devices.
environments. Moreover, Chatbots have access to a global
network of information via the internet and can be easily 3) Bridging Data across Platforms and Services
programmed to retrieve information in real-time which can IoT platforms can be seen as software development
improve the context. environments which handle Device management, Application
In practical terms, Chatbots simplify the way we consume management, Connection Management, Dashboard and
information from multiple screens and heavy data and graphics Analytics. Yet owing to platform fragmentation [17, 18],
to simple conversational interfaces capable of delivering highly sharing of data across platforms is still uncommon. One
contextual and intelligible information within the flow of the solution is to solve the issue at the application level by using
chat app itself. Achieving this high-level of abstraction can 3rd party services, which through APIs can access data from
deliver actionable intelligence (wisdom) with domain and user each platform. The data can be either processed on the various
knowledge to maximize the full potential of IoT. For example, platforms or extracted into another service and used to deliver
in use case (A), the query was relatively vague. The Chatbot something of value which can then be presented through a
could have used contextual information from Real time outside single Chatbot interface.
temperature along with knowledge from historical user
preferences to perform a specific action.
4) Search and Discoverability
b) Information Retrieval A key attribute in IoT is the natural tendency of objects
IoT dashboards are often saturated with various to be dispersed in the environment while being interconnected
metrics, data points, charts and tables making it difficult for and identifiable at class-level (i.e. common information across
users to find the required information. Chatbots can effectively the same class) or serial-level (i.e. unique to an individual
solve this problem by responding quickly to direct queries with object) [23].
highly accurate information. By understanding the specific Based on the permissions of the requester and the
intent of the user they limit the scope of information for availability of the connected objects in the scope of the
presentation. In terms of the knowledge hierarchy, Chatbots environment, IoT requires lookup and discovery services to
perform lookup and abstraction on IoT data. By automatically effectively find and control these objects. Such services include
providing IoT data as well as user-contextual data to an availability of sensors and actuators which the Chatbot would
analytics, Chatbots can also derive its own knowledge. For be able to retrieve from the entities and convey to the user at
example, in use case (B) the query only asked for Battery the appropriate times.
Charge related information and nothing else. The Chatbot
limited the response accordingly.
 As more use cases are discovered, chatbots can make the
5) Monitoring and Reporting adoption and diffusion of IoT systems significantly easier and
From IoT wearables such as health monitoring devices reduce the cognitive burden required to understand the
to industrial sensors which convey information in real time, functionalities of these systems.
monitoring and reporting are key aspects of IoT systems.
Chatbots can also be effectively used to perform as monitoring 2) User Interface Opportunities
services by integrating with solutions such as Application Graphical User Interfaces (GUI) for IoT are largely
Performance Management (APM). Accessing data from functional in nature. While it achieves simplicity by displaying
various IoT systems is a key advantage which is unique to virtual switches, sliders and buttons rather than passing
Chatbots in this scenario. Similarly Chatbot services can utilize complex commands, it still has some shortcomings which Chat
its own reporting services and present the abstracted interfaces can solve. (1) Chat interfaces understand natural
information to the user in an actionable and timely manner. language which makes interaction with the system as simple as
asking queries and receiving answers. There is no need for
navigation of menus and finding the right icon/button to
B. Human Centric Challenges of IoT
perform a task. (2) Chatbots use machine learning techniques
Chatbots were created with the primary purpose of to learn about an individual user and can personalize the service
improving the human-computer user experience. As such, to that user. In this way, they can understand the unique way
solving the user experience shortcomings of IoT systems can be the user converses with while maintaining the natural flow of
an important opportunity for chatbots. IoT, with its complex the conversation (3) They are also highly contextual interfaces
system of applications, sensors, actuators and services presents and can understand the intent in the scope of the past
a daunting challenge of gaining technical knowledge to interact interactions which is unique to chat based interfaces (also
with these various components. Hence exposing settings and speech). (4) Chat based interfaces concern mostly textual
configurations to users presents an obvious and unfriendly information thereby simple log files can be maintained and
burden that is far from ideal. consequently analyzed to make debugging easier.

1) Cognitive Burden 3) Configuration Challenges

Apart from the knowledge required to adapt to the new
The changing technology landscape of IoT is both imminent systems and ease cognitive burden, each IoT device has its own
and rapid. Furthermore, as newer features and use cases are unique setup and configuration in terms of software, network,
introduced, there is an added responsibility to educate the end firmware etc. As the number of different IoT devices increase,
users which can be burdensome for both the users and the it becomes difficult and burdensome at best to navigate the
developers of the system. Complicated systems cause interfaces of various applications and properly configure the
difficulties in adoption and diffusion. As an assistive system. Often technicians are involved to configure and explain
technology, chatbots can simplify the learning curve by the the uses of the system.
following ways: Using Chatbots, users can be guided and advised on the
a) Help Features: IoT enabled Chatbots can feature help right configurations for their system by creating step-by-step
texts which clarify the user request to ensure that the action setup processes using predefined configuration APIs. This also
performed is same as the one intended. reduces human effort in setting up the system. For example: A
b) Recommendations: Chatbots can recommend possible new device was configured in use case (D).
actions to the user which can be made more intelligent and
context aware depending on user preferences and the dynamics 4) Lack of Automated Error Reporting
of the environment. The distributed nature of most IoT systems implies that user
c) Automation: Chatbots are good at automating report databases of IoT errors are spread across multiple
common cyclic, tasks and can perform certain actions such as organizations, Operating System (OS) vendors, ISPs, and
monitoring availability of sensors (uptime, downtime etc.) and device vendors which makes automated problem reporting a
others through routine API calls, Websockets or Publisher- major challenge. Furthermore, users themselves are uncertain
Subscriber methods. which organization to report the particular issue to. Thus,
various stakeholders in the system have a limited understanding
d) Better Quality of Service (QoS): Feedback loops can
of the true nature of the problem and avoid sharing information
be easily integrated within chatbots to aggregate most frequent
with each other. Chatbots, in this scenario, can access these
queries and data from the process can be used to improve the
future Quality of Service (QoS). reported problems and by integrating other services, be able to
not only retrieve information from the IoT system but send
information to it. In Use Case (E), the chatbot identified the
correct stakeholder to send the error.
 A. IoT System
5) Support Challenges 1) IoT Devices
In the context of the paper, we refer to an IoT device
Remedying hardware and software issues in modern in the broad scope of the term as a “uniquely identifiable IoT
consumer IoT systems can be an irksome task. The recourse is endpoint which can be accessed and controlled using RESTful
to call the service provider for technical support or in many Web APIs”. In the situation an embedded device does not have
cases return the product. Either way it is an unnecessary burden APIs, there are existing solutions to easily create APIs for them.
on the user as well as the support vendors in today’s cost For Example: Using platforms such as Zetta, one can create IoT
structure. cloud based systems with full-fledged APIs. In the presented
Smart Chatbots often have support services built into their system we consider a Home automation system consisting of
functionality. It can even integrate human-in-the-loop systems Smart lights (eg. Philips Hue) and Smart HVAC (consisting of
to handle situations the Chatbot is not trained or authorized to a smart thermostat eg. Nest) as well as a Connected Car or
perform, in real-time. In this manner, users need not go beyond Smart Car (eg. Tesla). However, on principle, any IoT device
the scope of the chatbot application to look for product support. may be considered for interfacing with Chatbots.
Any software issue or hardware malfunction can be monitored
and Over the Air (OTA) software repairs can be performed. 2) IoT Cloud Platform
Chatbots can also be used to schedule technical repairs making IoT Cloud based Platforms is an important enabling
it a convenient and fast solution to customer support [24]. For technology in many IoT systems today. They deal with various
example, in Use case (E), a human operator was made to fragmented technologies in embedded devices from access
intervene. protocols (eg. Message Queuing Telemetry Transport (MQTT),
IV. SYSTEM DESIGN AND ARCHITECTURAL ELEMENTS HTTP etc.), wireless protocols (Zigbee, Bluetooth Low Energy
(BLE)) to various services, SDKs and integrations. There has
We present a conceptual system design which will aid
been positive reports on the established advantages of Cloud
in building Chatbot systems for IoT. Fig. 3 presents the high based IoT platforms [25]. Our system design stresses on
level view of the overall architecture consisting of the IoT
accessing and controlling the embedded devices in question,
system and the Chatbot system.
such as Smart Car, Light, Thermostat etc., through the API
Management/Gateway of the IoT Cloud Platform, regardless of
the standards and protocols of the individual embedded devices.
On the other hand, complex components in Networking and
Computing infrastructure for the Cloud platform have not been
addressed in this paper. Popular IoT platforms today include
Microsoft Azure IoT, IBM IoT, APIGEE IoT, Cisco IoT,
among others.
B. Chatbot System
1) Chatbot Channels and Platforms
Chatbot Channels are applications which run Chatbots
on supported Mobile devices (eg. Smartphones, Tablets) or
Terminals (eg. Desktop Applications).They are typically built
on top of the existing instant messaging platforms. Popular
Chatbot channels include Facebook Messenger, Slack,
Telegram, Kik, Skype, Line and Twilio SMS. These channels
are essentially the Chatbot applications in which a user interacts
with the bot. The area of Chatbot development is still in its
infancy and there can be many different architectural
approaches in implementing Chatbots.
In some approaches, the channels are interfaced
separately from the Chatbot Platforms through connectors.
Here, the Chatbot Platforms are hosted on cloud services which
can use Webhooks to communicate with the Channel. In the
context of this paper, we consider the integration of Chatbots as
text based inputs to IoT. By using Software Development Kits
(SDK) it is possible to integrate IoT to voice/speech based
Intelligent Personal Assistants such as Amazon Echo (using
Fig. 3. Proposed System Design of IoT-Chatbot System Alexa SDK) and Google Home (using Google Assistant SDK).
 c) Intents: Intents are the crux of conversational UI in
2) Chatbot Engine chatbots. The intents represent what the users are looking to
Perhaps the most important component of a Chatbot is the accomplish: get status updates, turn on/off devices, ask for help
engine, often referred to as Natural Language Understanding etc. The message passed from the user (utterance) in natural
(NLU) engine. It is responsible for translating natural language language is first analyzed for the intent. This means mapping a
into machine understandable action. Chatbot engines are often phrase to a specific action that should be taken by the IoT
highly complex, using various Natural Language Processing system as well as the specific dialog to be returned from the
(NLP) models and Machine Learning (ML) techniques to Chatbot. The information contained in an intent would be the
provide acceptable levels of accuracy. To make it easier for context and action.
Chatbot developers, many companies offer the processing
capability of the Chatbot engine as a Software-as-a- d) Action: Action refers to the steps that the IoT device
Service(SaaS) or ‘AI-as-a-service’ which are applied to will take when the intent of the user input is recognized. Actions
Chatbot applications using APIs. For example, Wit.ai and have specified parameters which categorize details about it.
Microsoft LUIS. Actions will be triggered once they are recognized by the intent
This paper is primarily focused on listing the relevant key For example the actions in a Smart Home may be
functional components of the engine in the context of IoT, not smartHome.lightsOn, smartHome.doorLock, and
on designing the NLP techniques for the Chatbot engine. Next, smartHome.getStatus. In this scenario, other parameters may
we include key concepts typically associated with chatbot also be defined such as location (eg. Dining room), time
engines [26, 27]: start/end (eg. 10am, Thursday etc.), schedule (eg. every hour,
every minute) etc.
a) Entities:
Entities are domain specific information extracted from Once the action has been set and the require parameters have
the utterance that maps the natural language phrases to their also been defined, the correct intent can be mapped to an IoT
canonical phrases in order to understand the intent. They help API endpoint and a HTTP request is made.
in identifying the parameters which are required to take a
specific action. To train the chatbot engine, entities which are Chatbots can also be built using existing frameworks
expected to give the same actions are typically grouped simplify the end-to-end process of creating and integrating
together. Common entities can be predefined as they can be Chatbots into messaging and IoT platforms. The key
used in many different scenarios. For example: Money, Color, advantages of using a Chatbot framework are: (1) Ease of
Date time, Location, Number. Domain specific entities can be Development from pre-defined actions, integrations and SDK
trained to recognize similar phrases. IoT devices are one such support for various IoT systems (2) Ability to ‘write once
domain specific entity. deploy anywhere’ through integrations with multiple Chatbot
For example, for the utterance: “Thermostat”, the Channels. (3) Ability to use AI-as-a-service. For example,
acceptable phrases may be trained to recognize "Thermostat", LUIS.ai in the case of Microsoft Bot Framework. Other popular
"heat", “heating”, “AC” ,"air conditioning" and will be Chatbot frameworks include API.ai, Microsoft Bot Framework
decoded as {“type”:iot, “device”: “Thermostat”} where the and IBM Watson Conversation Service.
entity is IoT. Entities may also have its own attributes. For
example, the Utterance: “$15” can be decoded as {“type”: V. FUTURE RESEARCH AREAS AND CONCLUSION
“money”, “amount”:15, “currency”: “dollars”} where the entity IoT is poised to become intrinsic to the day to day
is Money in JSON format. activities in the future, however the dynamic nature of IoT has
its share of difficulties, and this paper has put forward the
b) Context: concept of using Chatbots to address some of the challenges in
Determining the Context of the current user IoT. Through this initial endeavor we can identify possible
expression is an important feature of modern Chatbots. It can areas that can be worked on for the future which show great
be used to handle situations where the utterances may be vague potential:
and have multiple meanings depending upon the history of the 1) Stronger AI based Agents
conversation. Contexts represent the ability of agents to As more advances are made in the field of AI,
maintain state (also called lifespan or the number of utterances Software Agents will also grow to become more intelligent in
after which the context will be removed) and match the required the future. The goal of Strong AI has been to match the
intent. They may also use information from external machine’s intellectual capability to a human being. Immediate
sources. For example, if the user was asking about the living research challenges includes improving decision making ability
room (location) in the first utterance and then after the passing to create more autonomous Chatbots and better NLP as well as
of a few more utterances, mentions a vague statement such as: Natural Language Generation models to create more natural
“make it colder”, the Chatbot uses context to understand the flows of conversation between humans and bots. Chatbots will
earlier reference to the temperature of the living room. play an important role in the research areas of Intelligent Agents
as well as Machine-to-Machine (M2M) research in IoT.
 2) Cyber Physical Systems and IoT VI. REFERENCES
Cyber Physical Systems refer to more advanced, next
generation embedded Information and Communications [1] Atzori, L., Iera, A. and Morabito, G., 2010. The internet of things: A
Technology (ICT) systems. They share many similarities with survey.Computer networks, 54(15), pp.2787-2805 J. Clerk Maxwell, A
IoT but with higher combination and coordination between Treatise on Electricity and Magnetism, 3rd ed., vol. 2. Oxford: Clarendon,
1892, pp.68–73.
physical and computational elements [28]. The US National
[2] Evans, D. "The Internet of Things How the Next Evolution of the Internet
Science Foundation (NSF) identified cyber-physical systems is Changing Everything (April 2011)." (2012): 346-360.
among the key research areas in the foreseeable future [29]. IoT [3] van der Meulen, R., 2015. Gartner Says 6.4 Billion Connected ‘Things’
will play a major role in the transition to CPS as one of the key Will Be in Use in 2016, Up 30 Percent From 2015. Stamford, Conn
enabling technologies [16, 30]. Further advancements in AI [4] ITU-T Recommendation database", ITU, 2016. [Online]
aspects of Chatbots in IoT will be closely related to the Available:http://handle.itu.int/11.1002/1000/11559
Conversion, Cyber, Cognition and Configuration levels of the [5] Guinard, D., Trifa, V., Mattern, F., & Wilde, E. (2011). From the internet
of things to the web of things: Resource-oriented architecture and best
5C CPS Architecture [31]. practices. In Architecting the Internet of Things (pp. 97-129). Springer
Berlin Heidelberg.
3) Wisdom of Things [6] Guinard, Dominique; Vlad, Trifa (2015). Building the Web of Things.
The progress of Chatbots in IoT introduces the Manning. ISBN 9781617292682.
paradigm of human-in-the-loop systems which has exciting [7] Vermesan, O.,et al.., 2011. Internet of things strategic research roadmap.
research challenges in the areas of Wisdom of Crowds. The O. Vermesan, P. Friess, P. Guillemin, S. Gusmeroli, H. Sundmaeker, A.
Bassi, et al., Internet of Things: Global Technological and Societal
concept of wisdom of crowd suggests that aggregation of Trends, 1, pp.9-52
information can result in decisions that are better than what [8] Guinard, D., Ion, I. and Mayer, S., 2011, December. In search of an
could have been achieved by any individual in the group [32]. internet of things service architecture: REST or WS-*? A developers’
In the context of IoT, the sharing of big data from billions of perspective. International Conference on Mobile and Ubiquitous Systems:
Computing, Networking, and Services (pp. 326-337). Springer Berlin
sensors and devices creates more value in the ecosystem as Heidelberg
compared to not sharing. However it requires data [9] Evansdata.com. (2016). Evans Data Corporation | Internet of Things –
interoperability rather than simply accumulating multiple Vertical Research Service. Available at:
disparate data sources which are incompatible or have no http://www.evansdata.com/reports/viewRelease.php?reportID=38
similarities. Hence Chatbots in IoT systems can use techniques [10] Hewitt, C., 1977. Viewing control structures as patterns of passing
such as Human Swarming, an approach that uses real-time messages. Artificial intelligence, 8(3), pp.323-364
feedback loops from groups of users to make accurate insights. [11] Nwana, Hyacinth S. "Software agents: An overview." The knowledge
engineering review 11, no. 03 (1996): 205-244
There are plenty of interesting research opportunities in
[12] Schermer, Bart Willem. Software agents, surveillance, and the right to
acquiring accurate values from the crowd. privacy: a legislative framework for agent-enabled surveillance. Leiden
University Press, 2007
4) Evolution of the Semantic Web [13] Russell, Stuart Jonathan, Peter Norvig, John F. Canny, Jitendra M. Malik,
As the Internet itself changes, there are many more and Douglas D. Edwards. “Artificial intelligence: a modern approach”.
Vol. 2. Upper Saddle River: Prentice hall, 2003.
opportunities for exciting research in the areas of IoT and
[14] Broadband Commission, 2014. The state of broadband 2014: Broadband
Software Agents. The development of a Web 3.0 or Semantic for all. Geneva, Switzerland: The United Nations
Web and its impact on the future of Software Agents has been [15] S. Liang, "SensorThings API - connecting IoT devices, their location and
clearly described in the literature [33]. This poses a great theirdata,"2016.Available:http://www.eclipse.org/community/eclipse_ne
opportunity for research in IoT in terms of Semantic wsletter/2016/march/article2.php
interoperability which can have major impact on the IoT [16] Miorandi, D., Sicari, S., De Pellegrini, F. and Chlamtac, I., 2012. Internet
paradigm itself. The evolution of the Semantic Web will have of things: Vision, applications and research challenges. Ad Hoc
Networks, 10(7), pp.1497-1516
major impact on areas in pervasive computing, M2M
[17] Celesti, Antonio, Maria Fazio, Maurizio Giacobbe, Antonio Puliafito, and
technologies resulting in Software Agents being able to draw Massimo Villari. "Characterizing Cloud Federation in IoT." In 2016 30th
more value and achieve a higher level of wisdom than before. International Conference on Advanced Information Networking and
Applications Workshops (WAINA), pp. 93-98. IEEE, 2016
Development in the field of IoT has been phenomenal [18] M. Wallace, "Fragmentation is the enemy of the Internet of Things |
in recent times. Similarly, Chatbot systems are also becoming Qualcomm",Qualcomm, 2016
more intelligent and sophisticated as the days progress. To the [19] M. Littman and S. Kortchmar, "The path to A programmable world,"
2014.Available:http://footnote1.com/the-path-to-a-programmable-world/
best of our knowledge, no work has been published detailing
[20] W. Mckitterick, "The Messaging App Report: How instant Messaging can
the specific integration of Chatbots to IoT. This paper has be monetized," Business Insider
attempted to integrate these two fields together by enlisting the [21] Rowley, Jennifer E. "The wisdom hierarchy: representations of the DIKW
key architectural components required and envision possible hierarchy." Journal of information science (2007)
ways to address some of the present challenges in IoT. We hope [22] Barnaghi, P., Wang, W., Henson, C. and Taylor, K., 2012. Semantics for
that this endeavor will lead to more intelligent and effective the Internet of Things: early progress and back to the future. International
Journal on Semantic Web and Information Systems (IJSWIS), 8(1), pp.1-
integrated IoT systems. 21
[23] Bandyopadhyay, Debasis, and Jaydip Sen. "Internet of things: [29] Wolf, Wayne (November 2007). "The Good News and the Bad News
Applications and challenges in technology and standardization." Wireless (Embedded Computing Column".IEEE Computer. 40 (11): 104–
Personal Communications 58, no. 1 (2011): 49-69 105.doi:10.1109/MC.2007.404
[24] Vinyals, Oriol, and Quoc Le. "A neural conversational model." arXiv [30] Wan, J., Chen, M., Xia, F., Li, D. and Zhou, K., 2013. From machine-to-
preprint arXiv:1506.05869 (2015) machine communications towards cyber-physical systems. Comput. Sci.
[25] Google, "Overview of Internet of things," Google Developers, 2016. Inf. Syst., 10(3), pp.1105-1128
Available: https://cloud.google.com/solutions/iot-overview [31] Lee, J., Bagheri, B. and Kao, H.A., 2015. A cyber-physical systems
[26] Microsoft, "LUIS: Help," 2016. Available: https://www.luis.ai/Help architecture for industry 4.0-based manufacturing systems.
Manufacturing Letters, 3, pp.18-23
[27] API.ai,"Api.ai"2016. Available: https://docs.api.ai
[32] Joe Barkai. (2016). Wisdom of Things - Joe Barkai. Available at:
[28] Rad, C.R., Hancu, O., Takacs, I.A. and Olteanu, G., 2015. Smart http://joebarkai.com/wisdom-of-things
monitoring of potato crop: a cyber-physical system architecture model in
the field of precision agriculture. Agriculture and Agricultural Science [33] Berners-Lee, T., Hendler, J. and Lassila, O., 2001. The semantic web.
Procedia, 6, pp.73-79. Scientific american, 284(5), pp.28-37