Inferring User Search Goals with Feedback Sessions using K-means clustering algorithm

Abstract

Recognizing or inferring client's search objective from given query is a difficult job as search engines let users to identify queries simply as a list of keywords which might refer to broad topics, to technical terminology, or even to proper nouns that can be used to guide the search procedure to the significant compilation of documents. In order needs of users are correspond to by queries submitted to search engines and different users have different search goals for a broad topic. Sometimes queries may not exactly represent the user's information needs due to the use of short queries with uncertain terms. thus to get the best results it is necessary to capture different user search goals. These user goals are nothing but information on different aspects of a query that different users want to obtain. The judgment and analysis of user search goals can be improved by the relevant result obtained from search engine and user's feedback. Here, feedback sessions are used to discover different user search goals based on series of both clicked and unclicked URL's. The pseudo-documents are generated to better represent feedback sessions which can reflect the information need of user. With this the original search results are restructured and to evaluate the performance of restructured search results, classified average precision (CAP) is used. This evaluation is used as feedback to select the optimal user search goals.