Learning to rank deep web

With the fast advance of Deep Web, data sources selection according to database quality is crucial in Deep Web integration. In this paper, we exploit some features (such as capability of query interface, quality of interface pages etc) to learn to rank Deep Web through SVM and Ranking SVM. Otherwise, the feedback mechanism makes our ranking function more precision in ranking Deep Web. Experimental results show that our features are beneficial to estimate the quality of data sources, and the learned ranking function outperforms other existing data sources selection method which ranked the databases based on their desirability with respect to the query. Copyright ©2009 Binary Information Press.