Aggregating user-centered rankings to improve web search

This paper is to investigate rank aggregation based on multiple user-centered measures in the context of the web search. We introduce a set of techniques to combine ranking lists in order of user interests termed as a user profile. Moreover, based on the click-history data, a kind of taxonomic hierarchy automatically models the user profile which can include a variety of attributes of user interests. We mainly focus on the topics a user is interested in and the degrees of user interests in these topics. The primary goal of our work is to form a broadly acceptable ranking list, rather than that determined by an individual ranking measure. Experiment results on a real click-history data set show the effectiveness of our aggregation techniques to improve the web search.