Reexamination on Potential for Personalization in Web Search

摘要

Various strategies have been proposed to enhance web search through utiliz-ing individual user information. How-ever, considering the well acknowl-edged recurring queries and repetitive clicks among users, it is still an open issue whether using individual user in-formation is a proper direction of ef-forts in improving the web search. In this paper, we first quantitatively dem-onstrate that individual user informa-tion is more beneficial than common user information. Then we statistically compare the benefit of individual and common user information through Kappa statistic. Finally, we calculate potential for personalization to present an overview of what queries can bene-fit more from individual user informa-tion. All these analyses are conducted on both English AOL log and Chinese Sogou log, and a bilingual perspective statistics consistently confirms our findings.