Abstract:

A significant fraction of search queries relate to topics that are time dependent, where the ideal result changes from one time-span to the next. Despite their prevalance and importance, such queries have been largely ignored. In this paper, we focus on a subset of temporal queries known as implicitly year qualified queries. Such queries, which include ``olympics'' and ``miss universe'' implicitly depend on the year. We describe an efficient algorithm for automatically mining these queries using query log analysis. We apply our algorithm to a large query log consisting of over 750 million web search queries from a commercial search engine and analyze the characteristics of the mined queries. Finally, we show that web search effectiveness for certain classes of implicitly year qualified queries can be significantly improved using our proposed temporal result set reordering approach.