Semantic Retrieval Using Ontology and Document Refinement

Abstract—To enhance the retrieval accuracy of information search engine, this paper proposes a information retrieval system based on semantics and document refinement that realized by employing the semantic description and relevance of ontology to the information system. We describe the using of LSI (latent semantic indexing) approach to replace the traditional VSM (vector-space model) approach in detail in the results of sorting process and have a comparative experiment. Various experiments demonstrate the feasibility and effectiveness of our approach, LSI approach proposed in this paper is more effective and able to query the most relevant results in the top of the returnee for semantic retrieval than VSM and about 10.7%~22.2 % increase of the performance. Index Terms—ontology; document refinement; semantic retrieval; LSI; VS