Lexical features

Since version 2.18, a system of lexical features has been introduced to have a more fine-grained annotation of the status of the corresponding features. A lexical feature is an inherent feature associated with a lexeme, as opposed to a true morphosyntactic feature corresponding to an inflection of the morpheme.

Related publication:

Status of morphosyntactic features Illustration with written and spoken French UD treebanks (Kahane et al., TLT-SyntaxFest 2025)

The following lexical features are used (in French treebanks only for now):

• Gender[lex] for nouns;
• Number[lex] for pronouns such as moi;
• Person[lex] for pronouns such as moi.

Has been also been considered and could appear:

• Tense[lex] for auxiliaries such as En. will, Haitian pral, or Fr. aller.

Note that a word can have both Tense and Tense[lex], such as vais in je vais rentrer (which is Tense=Pres and Tense[lex]=Fut).

Note

These lexical feature are not kept in conversion to UD. Gender[lex]=Femis replaced in UD by Gender=Fem and Exponence[Gender]=Inherent.

Such a conversion would not be possible if a word has both both Tense and Tense[lex]. In such a case, Tense[lex] would be suppressed in UD.

See also UD issue 985.