Research On User Semantic Ontology Construction Based On User Log

In recent years, along with the rapid development of information technology, the numberof web pages on the Internet has become huge and is growing exponentially, the complexityand the short query of user entered and the ambiguity of the query of huge web pages havelead to a large challenge to search engine, how to meet people’s information need fast andreturn different results for different users. To provide personalization search results, we mustmine the user’s domain background knowledge. Most search engines gather a large number ofuser query logs, which record the user history queries and all clicked information, and theseinformation can reflect the user’s interest and domain knowledge to some extent. As ontologyis the key technology for semantic web, can provide domain terms and formal concepts, so itmade the interaction of information easy, and it can provide semantic background for searchengine.The main work in this paper are as follows:First of all, we present a novel similarity function to calculate the semantic similaritybetween user queries, using AGNES(Agglomerative Nesting) to cluster user queries, themethod can classify user query words into subjects by user’s personal interests andbackground knowledge. There are three semantic relationships between user queries based onuser query log, the query word itself, the expanded query and the clicked URLs. Then, byhierarchical agglomerative clustering algorithm, we can cluster user query terms into semanticsubjects and to disambiguate the semantic ambiguity of user query terms.Secondly, we propose a method to construct user semantic ontology. The steps are asfollows, according to the clustering results, form the original query and expanded queryformal context, optimize the expanded query formal context, then merge the original queryand optimized expanded query formal context and construct concept lattice, use the rule ofconcept lattice to ontology to form an initial ontology, finally, we use the normal ontologyWordNet to optimize the initial ontology. This ontology express the user interest preferences,and then applied to search engines, which will improve the technical level of informationcollection from based on similarity matching of keywords to based on semantic query.Finally, the experiments in this research employ VC++6.0program to prove the developapplications. And the experiment shows that based on our method, the user ontology canbetter reflect the latent semantic information of user query terms.