Learning to Rank with Voted Multiple Hyperplanes for Documents Retrieval
The central problem for many applications inInformation Retrieval is ranking.Learning to rank hasbeen considered as a promising approach foraddressing the issue.In this paper,we focus onapplying learning to rank to document retrieval,particularly the approach of using multiplehyperplanes to perform the task.Ranking SVM(RSVM)is a typical method of learning to rank.We point outthat although RSVM is advantageous,it still hasshortcomings.RSVM employs a single hyperplane inthe feature space as the model for ranking,which istoo simple to tackle complex ranking problems.In thispaper,we look at an alternative approach to RSVM,which we call Multiple Vote Ranker(MVR),andmake comparisons between the two approaches.MVRemploys several base rankers and uses the votestrategy for final ranking.We study the performance ofthe two methods with respect to several evaluationcriteria,and the experimental results on theOHSUMED dataset show that MVR outperformsRSVM,both in terms of quality of results and in termsof efficiency.
Learning to rank Document retrieval Ranking SVM Multiple vote ranker
He-li Sun Bo-qin Feng Jian-bin Huang
Department of Computer Science & Technology,Xian Jiaotong Univ.,Xian 710049,China School of Software,Xidian University,Xian 710071,China
国际会议
厦门
英文
572-577
2008-11-17（万方平台首次上网日期，不代表论文的发表时间）