Improved Accuracy and User Satisfaction by Inferring User Search Goals based on Feedback Sessions?

User search goals can be defined as information on various aspects of query that user want to obtain and it can be considered as the collection of information needs for a query. Different users may have different search goals in their mind when they pass ambiguous query to a search engine. Thus, it is important to infer and analyze user search goals to improve the performance of a search engine and user experience. By clustering the proposed feedback sessions, we infer different user search goals for a query. The feedback session is combination of both clicked and unclicked URLs and this feedback session is mapped to the pseudo-documents to better represent the information needs of user. These pseudo-documents are clustered using bisecting K-means clustering algorithm which produces better results than K-means clustering algorithm and reduces computation time. Finally, Classified Average Precision (CAP) evaluation criterion is used to evaluate the performance of system. In this way, the proposed system can infer user search goals efficiently and satisfy information needs of user. Experimental results are presented using user click-through logs from a commercial search engine to validate the effectiveness of our proposed methods.