UD for Laz 
Laz is a South Caucasian (Kartvelian) language that shares core structural features with Georgian—such as polypersonal verbal agreement, agglutination, and split-ergativity.
Tokenization and Word Segmentation
- Word Delimitation: In modern orthographies (both Latin-based and Georgian-based), words are regularly delimited by whitespaces and punctuation marks.
- Multi-word Tokens: Due to the highly agglutinative nature of Laz, certain grammatical clitics are attached directly to nominal or verbal bases and must be segmented into individual syntactic words:
- Auxiliaries / Copulas (AUX): Present-tense copula forms attach directly to nouns or adjectives. Example: noseri-on (she is smart) (\rightarrow ) noseri (ADJ) + on (AUX).
- Logophoric Particles (PART): The quotation particle -ya (that / they said) cliticizes to the very end of a cited clause. Example: on-ya (\rightarrow ) on (AUX) + ya (PART).
- Negative Particles (PART): The negative marker va- (not) frequently fuses with a following verb or copula. Example: va-it’urs (\rightarrow ) va (PART) + it’urs (VERB).
Morphology
Tags
- AUX (Auxiliary): Assigned to both independent past-tense copula stems (ort’u, t’u) and bound enklitic present-tense forms (on, yen, -re).
- PART (Particle): Used for negative particles (var, va) and the unique indirect/reported speech clitic ya (the Laz equivalent to Georgian -o).
- VERB: Assigned to content-bearing verbs. Masdars (verbal nouns) receive the VERB tag modified by the specific feature VerbForm=Vnoun.
Features
Nominal Features
- Laz completely lacks grammatical gender (Gender) and noun classes (NounClass).
- Nouns inflect for Number (singular vs. plural via the -pe/-ere suffixes) and Case.
- The case inventory includes: Nominative (typically unmarked), Ergative (-k), Dative (-s), Genitive (-iši), Instrumental (-ten), Allative (-ša), and Ablative (-din).
Verbal Features
- Verbs inflect for Tense, Aspect, and Mood.
- Polypersonal Agreement: A single Laz verb cross-references Person and Number not just for the subject (Person[subj]), but simultaneously for objects (Person[obj]) using a highly complex system of slot-based prefixes and suffixes.
Syntax
- Word Order: The default word order strongly leans toward SOV (Subject-Object-Verb), though it remains structurally flexible and pragmatically driven by topic and focus.
- Split-Ergativity: Case marking for core arguments (Subject nsubj and Direct Object obj) changes systematically depending on the tense layer (TAM Series):
- Present/Incomplete Tenses (Series I): Follows a Nominative-Accusative alignment. The subject of both transitive and intransitive verbs triggers the Nominative case, while direct and indirect objects trigger the Dative case.
- Past/Aorist Tenses (Series II): Follows an Ergative-Absolutive alignment. The subject of a transitive verb shifts to the Ergative case (-k), while the direct object shifts to the unmarked Nominative case.
Treebanks
There is one Laz UD treebank: