Empirical Exploitation of Click Data for Query-Type-Based Ranking

In machine-learning-based ranking, each category of queries can be applied with a specific ranking function, which is called query-type-based ranking. Such a divideand-conquer strategy can potentially provide better ranking function for each query categories. A critical problem for the query-type-based ranking is training data insufficiency, which may be solved by using the data extracted from click log. This paper empirically studies how to appropriately exploit click data to improve rank function learning in query-type-based ranking. The main contributions are 1) the exploration on the utilities of two promising approaches for click pair extraction; 2) the analysis of the role played by the noise information which inevitably appears in click data extraction; 3) the appropriate strategy for combining training data and click data; 4) the comparison of click data which are consistent and inconsistent with baseline function.