In this article, we discuss methodologies to extend computational semantic lexicons in a cost effective way using on-line language resources and savoir faire. First, we introduce the ecology of computational semantic lexicon acquisition, presenting two main methodologies: thesaurus-driven versus corpus-driven. Second, we describe an experiment to extend a semantics-based core lexicon with paradigmatic relations and predict the syntactic behavior of verbs based on their semantics; the automatically derived subcategorizations are first checked against corpora and then manually filtered. These lexicons have been developed within Mikrokosmos, a semantics-based machine translation system.
© 2001-2024 Fundación Dialnet · Todos los derechos reservados