Publication at Faculty of Mathematics and Physics |
2014
Abstract
This paper describes an experiment combining several existing data resources (parallel corpora, valency lexicon, morphological taggers, bilingual dictionary etc.) and exploiting them in a task of building a valency lexicon for a related language (Russian) derived from a high quality manually created valency lexicon for Czech (Vallex) containing several thousands of verbs with very rich syntactic and semantic information. The experiment is restricted only to nominal constituents in both simple and prepositional cases.
The results discussed in the second half of the paper seem to justify the method used and to encourage further experiments in this direction. The paper also discusses most frequent sources of errors.