Corpus-based Conceptual Characterisation of Verbal Predicate Structures

This paper describes performed research into the feasibility of creating computer-aided methods for knowledge extraction from an English corpus with the aid of existing lexical resources and analysis tools. Starting with subcategorised surface syntactic structures which give valuable information about the semantic and cognitive aspects of a (sub)language, the analysis is transferred to the abstract representation level of deep cases or thematic roles. These reveal information about the conceptual relations between the verbal arguments and the predicate that exist in the semantic domain under consideration. A semantic characterisation of the theta roles themselves will be supplied by using the lexical information of WordNet ((9]).