home edit page issue tracker

UD for Brahui

Tokenization and Word Segmentation

• In general, words are delimited by whitespace characters. Description of exceptions follows.
• According to typographical rules, many punctuation marks are attached to a neighboring word.

Morphology

Tags

• Unused/Absent UD Tags
  The following universal POS tags did not occur in the provided data and are therefore not used:

  • INTJ (interjection)
  • SYM (symbol)
  • X (other / unclassifiable)

• AUX vs VERB (how the distinction is drawn)
  The distinction follows UD guidelines with an English-oriented diagnostic:

  • AUX is used for verbs whose primary function is grammatical.
  • VERB is used for lexical verbs that provide the main predicate semantics.
    In ambiguous cases, the test used is whether the verb can be replaced by a full lexical verb without loss of core lexical content (if not, label AUX; otherwise, VERB).

• DET vs PRON (how the distinction is drawn)
  Following UD syntax-based diagnostics (English-style):

  • DET is assigned to words that serve to modify a noun and function as dependents inside a noun phrase (determiners, articles, demonstratives when they function as modifiers).
  • PRON is assigned to words that can act as the head of a noun phrase (personal pronouns, relative pronouns, reciprocals when they stand alone).
    Feature values (PronType, Person, Number, Case) are used to refine classification.

Features

• Note:
  The feature inventory and distinctions largely follow English-style analyses, adapted where possible to Brahui data.

• Nouns (NOUN)
  Nouns are annotated with features such as:

  • Gender: Not marked, as Brahui does not show grammatical gender in the available data.

  Features include:

  • Number=Sing|Plur
  • Case=Nom|Acc|Gen

• Verbs (VERB)
  Verbal features rely on UD conventions and English-based analysis where Brahui-specific descriptions were unavailable.

  Features include:

  • VerbForm=Fin|Inf|Conv
  • Tense=Past|Pres|Fut
  • Mood=Ind|Imp
  • Voice=Cau|Pass
  • Polarity=Neg
  • Person=1|2|3
  • Aspect=Imp|Perf

• Pronouns (PRON)
  Pronouns are annotated with features such as:

  • Gender: Not marked, as Brahui does not show grammatical gender in the available data.

  Features include:

  • PronType=Art|Dem|Ind|Prs|Rel|Tot
  • Number=Sing|Plur
  • Case=Nom|Acc|Gen
  • Reflex=yes
  • Poss=yes
  • Person=1|2|3

Syntax

Core Arguments (Subjects and Objects)

In this treebank, core arguments—subjects and objects—are annotated using rules largely based on English grammar.

Subjects (nsubj)

A token is annotated as a subject when:

• it represents the main agent or participant of the action,
• it typically precedes the verb, following English word-order conventions,
• it would function as the subject if the sentence were translated into English.

Subjects are consistently annotated across both verbal and nonverbal clauses.

Objects (obj, iobj)

Object relations follow straightforward principles:

• obj denotes the direct object, i.e., the entity directly acted upon by the verb.
• iobj denotes the indirect object, typically the recipient of an action.

In ambiguous cases, English grammar diagnostics were employed to assign the correct label.

Copula and Nonverbal Predicates

Some clauses lack a prototypical verb such as “run” or “eat” and instead express states or properties, such as being, becoming, or possession.

• Copula verbs that only show tense or time (like “is” or “was”) are labeled as aux.

Dependency Relations

The treebank employs the following Universal Dependencies (UD) relations:

acl, acl, advcl, advmod, amod, appos, aux,
case, cc, ccomp, compound, conj, det, flat, iobj, mark, nmod, nmod:poss, nsubj, nummod,
obj, obl, obl:tmod, parataxis, punct, root, xcomp.

Treebanks

There are 1 Brahui treebanks:

• Kholum

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

.