Handling polysemy in description logic ontologies

Abstract

Semantic Web technology highly depends on the quality of ontology as it reduces or eliminates conceptual confusion and reuses knowledge. In order to enhance quality of ontology, there is one major problem with lexical representation of ontology. Current lexical representation is term which may have different meanings, i.e., the term is polysemous, this can result in frustrating misunderstanding and ambiguity during the management and application of ontology. To solve this problem, sense is used to replace term as the lexical representation of concepts and properties for its unique meaning. The process of handling polysemy is to automatically disambiguate terms in ontology by using its surrounding ontology symbols and its nearby terms in annotated documents using this ontology. The right sense is assigned to a target term by maximizing the relatedness between the target and its neighbors. Experiments show our method has good performance. Comparing with the best word sense disambiguation method, the concept precision is almost 2 times than the precision of noun, and the property precision is almost 3 times than the precision of verb. The last experiment proves that our method is also effective in a semi-automatic process.