home edit page issue tracker

This page pertains to UD version 2.

It appears that you have Javascript disabled. Please consider enabling Javascript for this page to see the visualizations.

UD for Prakrit

Tokenization and Word Segmentation

Many Prakrit documents and inscriptions do not have whitespace to segment words. Attempt to follow the Sanskrit guidelines, including splitting transparent compounds and undoing sandhi.
Do not separate preverbs.

Morphology

Tags

All universal POS tags can be used (in theory).
Earlier inscriptions lack punctuation (no SYM) and auxiliary verbs are a late Prakrit innovation (no AUX).

Nominal features

Nominals (NOUN, PROPN, ADJ, some PRON, DET, NUM, and nominalised forms of VERB) have an inherent Gender feature with possible values Masc, Fem, or Neut.
Number on nominals has 3 possible values: Sing, Dual (rare) and Plur.
Case on nominals has 8 possible values: Nom, Gen, Dat, Acc, Loc, Abl, Ins, Voc.

Degree and Polarity

Polarity has two values, Pos and Neg, and applies to adverbs (ADV).

Verbal Features

Aspect is marked if it is perfective (Perf).
Mood has 2 possible values: Ind, Imp.
Voice has 4 possible values: Act, Pass, Mid, Cau (which combines with the others).

Pronouns, Determiners, Quantifiers

PronType is used with pronouns (PRON) and determiners (DET).
Person is a lexical feature of personal pronouns (PRON) and has three values, 1, 2 and 3.
- Person is also marked on finite verbs (VERB, AUX).

Other Features

There is one language-specific feature:
- Compound

Syntax

Special relations:
- acl:relcl for relative adnominal clauses.
- advmod:lmod and obl:lmod for adverbs/oblique nominals specifying a spatial relation.
- advmod:lmod and obl:tmod for adverbs/oblique nominals specifying a temporal relation.
- compound:redup for reduplication.
- nmod:desc for nominal epithets/regnal titles.

Treebanks

There is 1 Prakrit UD treebank:

Prakrit-DIPI (for imperial Prakrit inscriptions, including Ashokan edicts)