Learning to rank with minimal set of hyperplanes for document retrieval

Ranking is a central problem for web search, because the goodness of a search system is mainly evaluated by the accuracy of its ranking results. Learning to rank has been considered as a promising approach for ranking in Information Retrieval. In this paper, we focus on learning to rank for document retrieval, particularly using multiple hyperplanes based on order relationships to perform the task. Ranking SVM (RSVM) is a typical method of learning to rank. We point out that although RSVM is advantageous, it still has shortcomings. In this paper, we look at an alternative approach to RSVM and compared it with other state-of-the-art ranking techniques. Our approach uses the order relationship of the ranks to build the base decision functions and uses the vote strategy for final ranking. We study the performance of the ranking methods with respect to several evaluation criteria, and the experimental results on the OHSUMED dataset show that our approach outperforms other methods, both in terms of quality of results and in terms of efficiency.