Research Of Personalized Search Based On User Query Intent Identification

With the rapid growth of information on the Internet, user often gets massive search results when he/she uses the Search Engine, and most of the them are far away from user query intent,which cause serious information overload and users are disoriented at this time.People eager to have the Search Engine which can understand their personal information demand and return to these results which match user’s query intent well when user use the Search Engines.In view of this kind of situation, we propose a personalized search framework based on recognition of user query intent. The framework includes the following four parts.Firstly, the documents in the search corpus are preprocessed and a unitive probabilistic topic model named C-LDA is built.Secondly,computer uses C-LDA to solve probability distribution of user historical documents as his/her model.Furthermore, the user uses C-LDA and user model to identify his/her personalized query intent.Finally, the initial query will be expanded which is based on user intention and local co-occurrence algorithm, and the results will eventually be returned to the user.Innovation of this paper is as follows:(1) As to the existing poor precision to the current user modeling method,we propose to build user model with a public LDA model of search corpus. By comparison with the method that the topic model is constructed on user search history data directly, the NDCG is increased by 1.7%.(3)As to the existing poor ability to identify user query intent on query expansion,a new method for identifying user model is presented in this paper.Firstly computer use query topic and user model to identify user query intent,then original query is expanded which is based on the user’s query intent identification and term co-occurrence.By comparison with query expansion method based on independent component analysis semantic clustering, precision@ 5 is increased by 7.1% and precision@ 10 is increased by 7.4% and.