Effectiveness of the Information Retrieval techniques is becoming a great challenge due to enormous growth of web data. Information Retrieval precision can be improved if the information is retrieved using the information need of the user. This research proposes a method to improve the retrieval precision of search engine by using information scent and multimodal feature of clicked pages in query log mining. The algorithm is based on clustering query sessions using information scent of clicked URLs in the sessions which model the information need associated with the query sessions. In this approach search is driven by the information need of input query. Retrieval precision is improved by boosting the rank of clicked pages in retrieved multimodal web pages for the input query using the measure of similarity of input query to click pages in the selected cluster of query sessions. Performance of the proposed algorithm is evaluated with an experimental study of query log mining of all the Web search engine and it confirms the improvement in information retrieval precision.