engleski

CROVALLEX lexicon improvements: Subcategorization and semantic constraints

The paper describes the Croatian valence verb lexicon (CROVALLEX) that contains information on syntactic subcategorization and semantic restrictions of 1739 most frequent Croatian verbs. These 1739 verbs are associated with 5118 valence frames and enriched with 72 broad semantic classes with two further levels of subdivision (173 classes in total). The evaluation shows that syntacto-semantic verb classification helps in capturing the relation between the syntax and semantics of Croatian verbs and therefore reduces the redundancy in the lexicon. Unfortunately, classes in the current version of CROVALLEX do not provide a means for full inference of the verb semantics on the basis of its syntactic behavior. Therefore, in the improved version we plan to introduce the more distinctive semantic roles. In the improved version of CROVALLEX the semantic typing will be based on EuroWordNet Top Ontology. We believe that with such improvements we can solve the problem of sense differentiability and get a finer grained semantic classification of verbs in Croatian language.

Croatian verb valence lexicon; Valence frames; Syntacto-semantic classes; Verb synsets

nije evidentirano