Smoothing Clickthrough Data for Web Search Ranking

Wei Yuan (yuanwei <at> iro (point) umontreal (point) ca)

Le 17 juillet 2009 à 11 h 30 — jour inhabituel

Salle 3195, Pavillon André-Aisenstadt


Incorporating features extracted from clickthrough data (called clickthrough features) has been demonstrated to significantly improve the performance of ranking models for Web search applications. Such benefits, however, are severely limited by the data sparseness problem, i.e., many queries and documents have no or very few clicks. The ranker thus cannot rely strongly on clickthrough features for document ranking. We propose two smoothing methods to expand clickthrough data: query clustering via Random Walk on click graphs and a discounting method inspired by the Good-Turing estimator. Experiments on real-world data show that the ranking models trained on smoothed clickthrough features consistently outperform those trained on unsmoothed features. This study demonstrates both the importance and the benefits of dealing with the sparseness problem in clickthrough data.


Pour recevoir les annonces hebdomadaires par courriel, visitez http://rali.iro.umontreal.ca/rali/?q=fr/node/1631

Liste des autres séminaires