Learning to Cluster Feedback Session for Identification of User Search Objective

In web search applications, the WWW is a huge resource for people. This resource uses search engines to search the information. For this purpose the queries are submitted to search engine to represent the needs of the users. Sometimes queries may not exactly represent the actual objective of user. There may be ambiguous query for different users. There are mainly two aspects for improving search engine relevance 1) identification of user search objectives and 2) analysis of user search objectives. In this paper, we propose a new method to identify user search objective by analysis process. This process is operated on search engine query logs. First, we propose a skeleton for discovering different user search objective for a query by clustering the our proposed method of feedback sessions. Feedback sessions are calculated by considering user click-through logs. By using these user click-through logs the information needs of users can be efficiently identified. Second, we propose a new approach for generating pseudo-documents for best representation of the feedback sessions for hierarchical clustering. Hierarchical clustering gives the relationship between all the keywords in the corresponding cluster. Finally, we propose a new technique for actual evaluation of the performance of identified user search objectives. For the better effectiveness of our proposed methods results are presented using user click-through logs.