home v2/v2 edit page issue tracker

This page pertains to UD version 2.

Features in UD v2

We propose for v2:

The proposals in this chapter are based on

  1. experience with applying the v1 features (see also the issue tracker)
  2. survey of language-specific features and values used in current UD treebanks (especially those that are also documented)
  3. draft proposal from the UniMorph project, which has similar goals as UD features (Sylak-Glassman 2016; see below for a comparison of UniMorph and UD features)

Another reference that could contribute to the universal features is the GOLD ontology; see also the general project page. In particular, items in the ontology under Morphosemantic property and Morphosyntactic property roughly correspond to our features.

Renaming existing features and/or values

See the issue 219 for related discussion.

Adding/removing values to/from existing features

Adding new features

Comparison with UniMorph

General differences:

UniMorph dimensions (draft v2)

Stuff to check

Inventory of features that will stay language-specific

All layered features

We may want to standardize some of the layers but they seem to be de-facto standardized anyway.