Ontology-driven semantic ranking for natural language disambiguation in the OntoNL framework

Abstract

The measurement of the semantic relatedness has many applications in natural language processing, and many different measures have been proposed. Most of these measures use WordNet as their central resource and not domain ontologies of a particular context. We propose and evaluate a semantic relatedness measure for OWL domain ontologies that concludes to die semantic ranking of ontological, grammatically-related structures. This procedure is used to disambiguate in a particular domain of context and represent in an ontology query language, natural language expressions. The ontology query language that we use is the SPARQL. The construction of the queries is automated and also dependent on the semantic relatedness measurement of ontology concepts. The methodology has been successfully integrated into the OntoNL Framework, a natural language interface generator for knowledge repositories. The experimentations show a good performance in a number of OWL ontologies. © Springer-Verlag Berlin Heidelberg 2007.