UD for Gorani 
Gorani (ISO 639-3: hac) is a Northwestern Iranian language. In Universal Dependencies the first treebank documents the Hawrami (Hewramî) variety. The description below records the annotation decisions used for that treebank. They follow the published interlinear analysis of the source text and the universal guidelines. Where a Hewramî category does not yet have a settled UD analysis, the treebank keeps the universal label that the guidelines already define instead of inventing a language-specific one.
Tokenization and Word Segmentation
Tokens are orthographic words as printed in the source transcription. Punctuation is split off as a separate PUNCT token. There are no multiword tokens and no words with internal spaces. Clitics and other bound elements that the published interlinear gloss writes with = or - are still kept inside the orthographic word. The source gloss is preserved in MISC as Gloss. A later revision may split some of these clitics into syntactic words; until that review, person indexes, possessive clitics, additive particles, and copular material can remain attached to the host.
Sentence boundaries follow the numbered annotation units of the source recording. Those units are intonation units, so one UD sentence may contain more than one clause. The original unit id is kept in MISC as SourceUnit and in the sentence id (hwr-rec-zb-001 corresponds to ZB.1). If a token is not followed by a space in the sentence text, its MISC contains SpaceAfter=No.
Morphology
Tags
The treebank uses universal POS tags only. There is no language-specific XPOS tagset (XPOS is _).
NOUN,PROPN,ADJ,ADV,NUM,PRON,DET,ADP,CCONJ,PART,INTJ,VERB,AUX,PUNCT, andXoccur in the data.AUXis reserved for forms of the verb “to be” (lemmasbîandbîyê). Other verbs, including lexical “come” and “take”, areVERB.DETis used for demonstratives and forherçê“whatever” when it determines a nominal or a clause. Independent personal pronouns arePRON.PARTis used for the emphatic particleher.Xis used only when the conversion cannot assign a defensible universal tag. It should not be resolved by analogy with another Iranian language.ADPcovers both prepositions and elements that the source treats as postpositions when they are written as separate words.
Features
Only universal features are used. Every value below is the UD value, not the label of the source gloss.
GenderisMascorFem. It is marked on nouns, adjectives, pronouns, and on verbs or auxiliaries when the source glosses a gender agreement marker.NumberisSingorPlur.Personis1,2, or3on finite verbs and on auxiliaries that index a person.Casehas two values. The Hewramî direct case isCase=Nomand the oblique case isCase=Acc, which is the universal mapping for a two-case direct/oblique system. These values occur on nouns, pronouns, proper names, and on some adverbs and adpositions whose source form is an oblique pronoun or noun.DefiniteisDeforIndon nouns, andIndalso occurs on a few adjectives and adverbs whose source gloss marks an indefinite ending.Degree=Cmpmarks the comparative.PronTypeisDemfor demonstratives andIntfor interrogative or “whatever” forms.- Verbal features are
VerbForm=FinorVerbForm=Part,Mood=IndorMood=Sub,Tense=PresorTense=Past,Voice=Pass, andPolarity=Neg. A participle that also contains an unsplit present auxiliary keeps a singleTensevalue,Past, because a token cannot carry two tenses. The present piece remains visible inGloss.
Syntax
The dependency trees are a first conversion and are not yet a full manual review. Core arguments use nsubj and obj. Other nominal dependents of a predicate use obl, with the adposition attached by case. Coordination uses conj and cc. Discourse particles use discourse. Punctuation is attached with punct.
The verb “to be” is tagged AUX and, in this first conversion, is usually attached with aux. Where the source unit is a nonverbal clause and no lexical verb is present, the auxiliary may be the root and the nonverbal predicate may be attached with dep. The relation dep is also used where the automatic conversion did not yet choose a more specific label, especially for ezafe or linker constructions, complex predicates, and reported speech. Those dep edges are placeholders for later review, not a proposed Hawrami relation subtype. No language-specific relation subtypes are used.
Treebanks
There is one Gorani UD treebank: