Ontology-Based Approach for Semi-automatic Generation of Subcategorization Frames for Spanish Verbs

Abstract

This work deals with the semi-automatic generation of subcategorization frames (SCFs) of Spanish verbs; specifically, given a set of verbs in Spanish and their respective sense, their SCFs are obtained. The acquisition of SCFs in Spanish has been approached in different works: in some the frames are generated manually, while in others they are obtained semi-automatically from a tagged corpus; unfortunately in this case, the results depend on the characteristics of the texts used. The method proposed in this document combines an ontology-based approach (through lexical relations of verbs) and linguistic knowledge (functional class of verbs). The relations among base verbs and other verbs were obtained from the Spanish WordNet ontology, which contains lexical relations among words. Also, the existing relation between the SCF and the functional class of verbs was used to generate the SCFs. In order to evaluate the method the SCFs for 44 base verbs were generated manually, from which 239 SCFs were semi-automatically generated and validated, yielding an accuracy of 89.38%.KeywordsLinguistic KnowledgeTransitive VerbDependency ParsingCorrect CombinationDirect ComplementThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.