基于PairWise排序学习算法研究综述

Learning to Rank(L2R)技术是对搜索结果进行排序,是近几年的研究热点。现关于L2R中的PairWise方法进行研究分析,PairWise方法将排序问题转化为二元分类问题,其缺点是只考虑两篇文档的相对顺序,而不考虑文档出现在搜索结果列表中的位置。另外,不同的查询拥有的文档对数目不同,结果会向拥有文档对较多的查询偏移。对常用的PairWise算法的损失函数及其求解方法、基本思想、算法框架、效用评价以及算法应用进行概括分析。

Survey on PairWise of The Learning to Rank

Learning to Rank is useful for search results in a ranking task,which is a hot topic in recent years.The PairWise in L2R is studied and analyzed in this paper.The PairWise approach is transformed into binary classification problem in ranking of documents,but only relative order of two documents is considered,and the positions of the search result list is ignored in the documents.In addition,different queries have different numbers of document pairs, and the result is offset to more queries of document pairs.This paper summarizes the loss function and its solution methon,compares basic idea and algorithm framework,analyzes evaluation metrics and their applications of the commonly used PairWise algorithm.