home edit page issue tracker

UD for Dargwa

Tokenization and Word Segmentation

• Words are generally delimited by whitespace.
• Punctuation marks are tokenized as separate tokens.
• There are no multiword tokens written with whitespace.
• Imperatives used with particles are tokenized as one token.

Morphology

Tags

• Dargwa-Mehweb treebank currently uses 14 part-of-speech tags, with the exception of ADP, SCONJ, SYM.
• The tag PART is applied to the clitics ca (proclitic used in irreal clauses), cula (focusing particle), gʷa (the so called assertive copula, focus particle), hara(warning particle), ile (complementizer), k’a (topicalization marker), k’ʷan (quotative particle), ʡaj (particle following apprehensives).
• The tag AUX is applied to four lexemes: le (copula in non-verbal predication), ašes (hortative periphrastic mood marker), es (jussive periphrastic mood marker), uʔes (irrealis periphrastic mood marker).
• Participles are tagged as a VERB.

Features

Nominal features

• Nouns, pronouns and determiners are tagged with the Case feature, and are annotated for Gender and Number agreement.
  • Dargwa has three types of case markers: functional, orientation, and localization. Locative forms can be annotated with the combination of orientation and a localization case features.
  • Class agreement is tagged with the Gender feature, which have six values. Masculine, two feminine and neuter values are used with singular forms; human and non-human values are used with plural forms.
• Numerals have the Case feature.

Verbal Features

• Verbs and auxiliaries are tagged with Tense, Aspect, and Mood features.
  • There can be more than one mood markers, such forms are tagged with two values, e.g. Mood=Cnd,Opt.
• Verbs and auxiliaries are annotated for VerbForm: Conv, Inf, Part, Fin.
• Verbs and auxiliaries are annotated for negative Polarity.
• Verbs and auxiliaries agree for Gender and Number on an ergative basis.
• Egophoric agreement (with 1st person subject in affirmative contexts and with 2nd person subject in interrogative contexts, person agreement works on an ergative basis) is tagged with the Personsubj feature.
• There are two language-specific verbal features: ConvType and Evident.
  • ConvType is used with converbs and participles.
  • An evidentiality feature Evident is used with finite verbs to refer to the fact that the speaker has indirect evidence for the statement (non-first hand sourse of information).

Lexical (classification features)

• NameType is used for proper nouns (PROPN).
• PronType is used with pronouns (PRON) and determiners (DET).
• NumType and NumForm are used with numerals (NUM).
• Reflex is used with pronouns (PRON).
• Person is a lexical feature of personal pronouns (PRON) and has two values, 1 and 2.

Syntax

• Dargwa has ergative-absolutive case alignment, ergative is marked, while absolutive is unmarked. Ergative-marked arguments are semantic subjects of transitive verbs (A arguments). Absolutive arguments are semantic subjects of intransitive verbs (S arguments), semantic objects of transitive verbs (P arguments), and semantic subject of transitive verbs (A arguments) with certain temporal-aspectual forms (in the so-called bi-absolutive construction). Predicates agree with the absolutive argument; the case of the argument, the agreement pattern, and the valence of the predicate are used to distinguish between the subject and the object argument.

• There two language-specific relations: nsubj:nc (for a non-canonical nominal subject) and obl:float (oblique as a floating modifier).

• The subtype relations used are as follows:

  • compound:lvc
  • compound:redup
  • compound:svc
  • det:poss
  • flat:foreign
  • flat:name
  • obl:tmod

Treebanks

There is one Dargwa UD treebank:

• Dargwa-Mehweb

© 2014–2024 Universal Dependencies contributors. Site powered by Annodoc and brat

.