基于LambdaMART的个性化搜索检索模型  

Personalized Search and Retrieval Model Based on LambdaMART

作　　者：金众威 刘淑芬[1] 包铁[1] 

机构地区：[1]吉林大学计算机科学与技术学院,长春130012

出　　处：《吉林大学学报（理学版）》2016年第4期821-826,共6页Journal of Jilin University:Science Edition

基　　金：国家自然科学基金(批准号:61472160);国家科技支撑计划项目(批准号:2014BAH29F03)

摘　　要：针对如何把个性化信息加入到搜索结果排序中,提出一种基于决策树的可量化用户个性化信息的方法,并根据用户的搜索关键词与用户的个性化信息,预测用户的搜索意图,把预测结果融合在排序结果中,解决了传统检索模型无法有效加入用户个性化信息的缺陷.实验结果表明,加入个性化信息后的排序结果准确性明显提升,从而改善了用户对搜索引擎的体验.In view of the problem that how to add the personalized information to the search result ranking, we proposed a method based on decision tree to quantify the user＇s personalized information. According to the user＇s search keywords and user＇s personalized information, the method predicted the user＇s search intention, and fused the predicted results in ranking results. It can effectively solve the defect of traditional retrieval model by adding user＇s personalization information. The experimental results show that accuracy of the ranking results is significantly improved after adding the personalized information, thereby improving the user＇s experience of the search engine.

关 键 词：决策树 增强树 搜索引擎 个性化搜索 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]