Query clustering for boosting web page ranking

Abstract

Over the past few years, there has been a great deal of research on the use of content and links of Web to improve the quality of Web page rankings returned by search engines. However, few formal approaches have considered the use of search engine logs to improve the rankings. In this paper we propose a ranking algorithm that uses the logs of search engines to boost their retrieval quality. The relevance of Web pages is estimated using the historical preferences of users that appear in the logs. The algorithm is based on a clustering process in which groups of semantically similar queries are identified. The method proposed is simple, has low computational cost, and we show with experiments that achieves good results.