A distributional similarity measure for query-dependent ranking in web mining

Ranking model construction is an important topic in information retrieval and web mining. Recently, many approaches based on the idea of "learning to rank" have been proposed for this task and most of them attempt to score all documents of different queries by resorting to a single function. In this paper, we propose a distributional similarity measure for query-dependent ranking. In the query-dependent ranking framework, an individual ranking model is constructed for each training query with associated documents. When a new query is asked, the documents retrieved for the new query are ranked according to the scores determined by a joint ranking model which is combined from the individual models of similar training queries. The distributional similarity measure is used to calculate the similarities between queries. Experimental results show that our method is more effective than other approaches.