UD for Zazaki 
Tokenization and Word Segmentation
The current Zazaki-ZSD treebank (200 sentences, 1440 syntactic words) contains 70 multiword tokens that are split into multiple syntactic words. Typical patterns are:
- Noun + Ezafe, e.g. çarê → çar
NOUN+ êADP, dewa → dewNOUN+ aADPwith a following possessor (dewa ma “our village”), adjective (miletê kanî “the old people”), orMARKin relative constructions. - Pronoun + directional particle in preverbs, e.g. pi-ro → pi
PRON+ roADP, with ro annotated as a verbal particle (compound:prt).
Clitics are segmented as separate tokens and marked with full morphological features (Person, Number, Gender). Morphological affixes for case remain attached to their hosts.
Adpositions remain host-attached but are written seperately.
Morphology
Tags
Out of the 17 universal POS tags, all except SYM occur in the Zazaki-ZSD treebank. The following tags are attested:
ADJ, ADP, ADV, AUX, CCONJ, DET, INTJ, NOUN, NUM, PART, PRON, PROPN, PUNCT, SCONJ, VERB, X.
The most frequent categories are PUNCT (312), VERB (241), NOUN (216), PRON (196), and ADP (186), reflecting the spoken, narrative nature of the corpus.
AUX vs. VERB
The tag AUX (40) is used exclusively for forms of the copular and auxiliary verb bîyene “to be”, both in copular clauses and in periphrastic constructions expressing existence (e.g., est o). All other verbal lexemes, including light verbs (kerdene, vatene, kutene etc.), are tagged as VERB (241).
ADP
ADP may refer to proper pre-, post- and circumpositions like de, ra, bi (adposition type is not distinguished in the data), or to Ezafe particles in possessive constructions that are attached to the head noun with the relation Case. In the latter case, there is a distinction of Gender, Number, and Case.
PART
The tag PART (49 instances) is used mainly for discourse and pragmatic particles typical of spoken data, including fillers, attention markers, and pragmatic modifiers such as yanî, îşte, ya, and hanî. These particles are normally attached via discourse.
Other uses include semantically devoid preverbal elements of complex verbs, such as arê and hes (3 instances each). These particles are attached via compound:prt.
INTJ
INTJ (27 instances) is used for independent interjections, often occurring at clause boundaries or in repairs.
X and foreign material
X is used in only 2 instances. It is restricted to clearly foreign lexical items. Foreign material is indicated by the feature Foreign=Yes. Foreign sequences are occasionally structured using flat:foreign.
Features
The following universal morphological features are attested in the Zazaki-ZSD treebank:
Nominal domain
Gender:Masc,FemNumber:Sing,Plur-
Case:Nom,Acc
Since Zazakî has a bicasual system,Nom(24) stands for the direct case, whereasAcc(77) describes the oblique case. Case is only annotated if it is morphologically marked.Case marking is mostly restricted to pronouns and a small number of nouns, determiners, and proper names.
Definite:Ind
Pronominal features
Person:1,2,3Number:Sing,PlurGender:Masc,FemPronType:Prs,Dem,Int,Ind,RcpReflex:Yes(for the reflexive pronouns xo)Animacy:AnimAnim(2 instances) is restricted to the d-form pronouns of Northern Zazakî.
Verbal domain
Tense:Pres,PastMood:Ind,Sub,ImpAspect:Imp,PerfPolarity:NegVerbForm:Fin,PartVoice:Pass(only in one instance)
Other features
NumType:Card(for numerals)Foreign:Yes(for foreign or code-switched material)
No language-specific features are introduced. Several features occur only marginally and should not be interpreted as fully systematic morphological categories in this version of the treebank.
Syntax
Core arguments
Subjects of finite clauses are annotated with nsubj, regardless of morphological case marking. Zazakî exhibits predominantly SOV word order. However, as is expected, goals and recipients typically appear post-verbally. In addition, the spoken data shows some instances of afterthought or discourse particles that appear after the verb.
Direct objects are annotated with obj, while indirect objects (iobj) occur only rarely with recipients. Oblique arguments and adjuncts are annotated with obl, often accompanied by an adposition marked with the relation case.
Nominal modification and possession
Possessive constructions are annotated with the relation nmod:poss. These constructions commonly correspond to Ezafe structures, where the Ezafe marker is annotated as ADP with the relation case attaching it to the head noun.
Other nominal dependents are annotated using standard relations such as amod, det, nummod, nmod, and appos.
Reduplicated forms are annotated with compound:redup.
Light verb and particle constructions
Light verb constructions are a productive syntactic pattern and are annotated using compound:lvc, linking a nominal element to a verbal head.
Verbal particles and preverbs (e.g. pi-ro) are annotated with compound:prt.
Asyndetic forms are annoted with conj.
Clause structure
Copular clauses use the relation cop, with the non-verbal predicate as the clause head and the copular auxiliary (bîyene) attached as a dependent.
Finite clausal complements are annotated with ccomp, while non-finite or subject-controlled complements are annotated with xcomp.
Adverbial clauses are annotated with advcl and are typically introduced by subordinators (ke) marked with the relation mark.
Relative clauses are annotated with acl:relcl.
Discourse and spoken-language phenomena
Given the conversational nature of the data, the treebank contains a rich set of discourse-related constructions. Discourse particles and fillers are attached with the relation discourse (74 instances).
Repairs and false starts are annotated using reparandum (4 instances). Independent or loosely connected clauses are linked via parataxis (7).
Left- and right-dislocated constituents are annotated with dislocated (6 instnaces), and direct forms of address use the relation vocative (5).
Coordination
Coordination is annotated using conj for conjuncts and asyndetic verbs, and cc for coordinating conjunctions. Punctuation is attached using punct to the most relevant syntactic head.
Treebanks
There is 1 Zazaki UD treebank: