Automatic Discovery of Semantic Relations Based on Association Rule

Automatic discovery of semantic relations between resources is a key issue in Web-based intelligent applications such as document understanding and Web services. This paper explores how to automatically discover the latent semantic relations and their properties based on the existing association rules. Through building semantic matrix by the association rules, four semantic relations can be extracted using union and intersection in set theory. By building a cyclic graph model, the transitive path of association relation is discovered. Document-level keywords and domain-level keywords as well as their parameters are analyzed to improve the discovery accuracy. Rules can be gained from the experiments to optimize the discovery processes for relations and properties. Further experiments validate the effectiveness and efficiency of the relation discovery algorithms, which can be applied in Web search, intelligent browsing and Web service composition.