UD for Ancient Hebrew 
Tokenization and Word Segmentation
No tokens in the Ancient Hebrew treebank should contain whitespace. The following are treated as clitics and made into separate tokens:
- Prepositions (ב, ל, מ, …)
- Possessive and object pronouns (ני, נו, ו, ם, …)
- The corresponding independent pronoun is used as the lemma
- Conjunction ו
- Definite determiner ה
- This includes ה when it appears as demonstrative agreement on adjectives, participles, and demonstrative determiners
- Since the text includes vowels diacritics, ה is included as a token even when it does correspond to a full character in the consonantal text.
Morphology
Tags
All tags are used except X and SYM. AUX is used for the copula היה.
The positive and negative existentials ישׁ and אין are tagged VERB.
Participles are tagged either VERB or NOUN. If they have arguments or obliques, they are tagged as VERB, but if they do not then they are tagged as NOUN if they participate in nominal phrases.
Verbs in the infinitive absolute which are used for emphasis are currently tagged as ADV and attached to the inflected verb with advmod.
Features
The following universal features are in use:
- Aspect: AUX, VERB
- Gender: ADJ, AUX, NOUN, PRON, VERB
- Mood: VERB
- Number: NOUN, PRON, ADJ, VERB
- Person: VERB
- Tense: VERB
- VerbForm: VERB
The following language-specific features are in use:
The following MISC features are present:
SpaceAfter=No
Syntax
The subtypes compound:smixut and nmod:poss are used. The relation compound is currently unused.
The relations iobj, expl, and clf are unused.
The relations fixed, list, orphan, goeswith, reparandum, and dep are currently unused, but may be used in future.
Treebanks
There is 1 Ancient Hebrew UD treebank: