Research On Concept Similarity Of Web Information Retrieval

With the rapid increment of the information on the web, information retrieval is becoming more and more important. How to improve the efficiency of information retrieval becomes a hot research point. In the field of information retrieval, concept similary plays an import part, it will be provided with theory significance and practical significance to improve concept similarity precision for improving the efficiency of information retrieval.Presently, in the field of information retrieval, most methods of concept similarity compute similarity based on the surface of concept, only with exterior forms, without native meanings, and this made the improvement of concept similarity precision become a bottle-neck. While Ontology of Semantic Web can describe the semantic of information, it provides a new idea for disposing the problem. So the paper improved bottom concept similarity method and upper concept similarity method based on the deficiency of existing methods.Due to existing concept similarity methods limiting to one-layer information of concept, which can not reflect the degree of concept similarity, according to the improvements of literal and semantic and pragmatic similarity methods of this paper, a method based on multi-layers information was proposed. With the method, the bottom concept similarities were computed, and the results were compared with subjective judgments, the method is proved that it can improve the precision of bottom concept similarity.Based on the difference between bottom concept and upper concept, as well as the character that the meaning of ontology bottom concept are clearer than the meaning of upper concept, the paper analyzed the deficiency of the other methods which compute upper concept similairty, imported two coefficient which are the difference of layer-depth and the number of instances, put forward a method of computing upper concept similarity based on bottom concept similarity. This method can compute the similarity of two ontologies. The experimental results were compared with the results of classic probability statistic; it is provided that the method can not only improve the concept similarity precision, but also compute the similarity of ontologies.Finally, an information retrieval archetype based on ontology was proposedaccording to above methods, the functions of all the models which made up of the archetype were described.