Intelligent Information Retrieval Based On Ontology

With the development of network technology and its broad application, information retrieval has been paid more and more attention. The traditional information retrieval methods concentrate on the way of key-word matching. Semantic understanding, the most important part of information retrieval system has not been involved. Although recall The retrieval results can not meet the users' need.Inteligent information retrieval could process the resources semanticly compared with troditional information retrieval. In inteligent information retrieval, ontology is the foundation of semantic reasoning. Domain ontologies define concepts of the certain domains, describe relations of concepts and provide the logic rules for semantic reasoning. In this way computer could understand the structure and metadata of a domain ontology and get the obvious or recessive knowledge from information base by related logic rules. In this paper, we integrated Ontology and user relevance feedback into information retrieval and did some something based on the current information retrieval theory.1. A new intelligent information retrieval model was proposed. It consists of knowedge base, query expansion and concepts filtration. Applying Ontology to information retrieval is impossible to solve the problem of semantic understand insufficiency in a way.2. A new modeling method of domain ontology is presented based on the existing models of domain ontologies. The proposed method performs well in logic and operation.3. In order to improve the performance of information retrieval systems, a novel method for query expansion is presented. The proposed method is a hybrid QE technology that combines relevance feedback and ontologies. FirteX, which is the first open source information retrieval experimental platform in our country is used as our experimental platform. We compared the proposed method with cosine similarity-based QE which is a widely used query expansion technique. The experimental results show that the proposed method performs well.