Ontology Mining for Semantic Interpretation of Information Needs

Abstract

Ontology is an important technique for semantic interpretation. However, the most existing ontologies are simple computational models based on only "super-" and "sub-class" relationships. In this paper, a computational model is presented for ontology mining, which analyzes the semantic relations of "part-of", "kind-of" and "related-to", and interprets the semantics of individual information need. The model is evaluated by comparing the knowledge mined by it, against the knowledge generated manually by linguists. The proposed model enhances Web information gathering from keyword-based to subject(concept)-based. It is a new contribution to knowledge engineering and management.