Web search result optimization by mining the search engine query logs

Modern Information Retrieval Systems match the terms of a user query with available documents in their index and return a large number of Web pages generally in the form of a ranked list. It becomes almost impractical at the user end to examine every returned document, thus necessitating the need to look for some means of result optimization. In this paper, a novel result optimization technique based on learning from historical query logs is being proposed, which predicts users´ information needs and reduces their navigation time within the result list. The method first performs query clustering in query logs based on a novel similarity function and then captures the sequential patterns of clicked web pages in each cluster using a sequential pattern mining algorithm. Finally, search result list is re-ranked by updating the existing PageRank values of pages using the discovered sequential patterns. The proposed work results in reduced search space as user intended pages tend to move upwards in the result list.