Semi-automatic Verb Valence Frame Assignment through VerbNet Classification

An approach to cross-linguistic transfer of verb valence frames for Czech and Croatian language using VerbNet syntactic-semantic classification is presented. Confronting Czech-Croatian pairs of valence frames for each syntactic-semantic class, we extracted the semantic part of the Czech valence frames and applied the subcategorization features to Croatian with minor (mainly syntactic) modifications. The encouraging results suggest that other similar (Slavic) languages could benefit from the same approach. It is shown that even for languages that do not have sufficiently developed NLP resources that are highly relevant for verb subcategorization (such as WordNet), it is possible to obtain valence lexicon applying small number of rules for cross-linguistic syntactic transfer and using Levin’s verb classes.