Study On Automatic Identification Of User Semantic Intention For Personalized Search

Currently, most search engines apply the technique of keywords matching, i.e. the same results will be returned for the same query keyword. But in fact, the same keyword issued by different users may have the different query intention. Even the same keyword issued by the same user in different time may have different query intention. The personalized search technique appears for providing the personalized result for the personalized information requirement from different users and solving above problem. The research on polysemy of query words is one of the good approaches for implementing personalized search.Polysemy of query words decreases the accuracy of search results. To solve the problem, this thesis proposes the idea of automaticaly identificating user semantic intention when the user issues a polysemy query word. Here user semantic intention means the real meaning of query terms that user wants to express in current session. A key technique of identifying user semantic intention is word sense disambiguation based on web environment. This thesis puts emphasis on studying context-based query disambiguation and word sense disambiguation in contextless queriesis, and discussing the technique of query expansion based on user semantic intention.Word sense disambiguation in contextless queries is an innovation of the thesis. A novel solution, Implicit Context Based Word Sense Disambiguation (ICBWSD) is presented to solve the problem that existing methods can't deal with the single word disambiguation. This approach obtains the implicit context of query term by recording the immediate preceding query and the clicked documents information on the current query. After obtaining the implicit context, context-based method is utilized to determine the sense of the target word.Experiments show ICBWSD is feasible and promising in inferring user's query intention.