Disambiguation of polysemous verbs for rule-based inferencing

  • Authors:
  • Fabienne Martin;Dennis Spohr;Achim Stein

  • Affiliations:
  • Universität Stuttgart, Germany;Universität Stuttgart, Germany;Universität Stuttgart, Germany

  • Venue:
  • IWCS-8 '09 Proceedings of the Eighth International Conference on Computational Semantics
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

We present an approach to disambiguating verb senses which differ w.r.t. the inferences they allow. It combines standard ontological tools and formalisms with a formal semantic analysis and is hence more formalised and more detailed than existing lexical semantic resources like WordNet and FrameNet [Fellbaum, 1998, Baker et al., 1998]. The resource presented here implements formal semantic descriptions of verbs in the Web Ontology Language (OWL) and exploits its reasoning potential based on Description Logics (DL) for the disambiguation of verbs in context, since before the correct sense of a verb can be reliably determined, its syntactic arguments have to be disambiguated first. We present details on this process, which is based on a mapping from the French Euro WordNet [Vossen, 1998] to SUMO [Niles and Pease, 2003]. Moreover, we focus on the selectional restrictions of verbs w.r.t. the ontological type of their arguments, as well as their representation as necessary and sufficient conditions in the TBox. After a DL reasoner has identified the verb sense on the basis of these conditions, we make use of the more expressive Semantic Web Rule Language to calculate the inferences that are permitted on the selected interpretation.