吉林大学学报(理学版)

• 计算机科学 • 上一篇    下一篇

基于LambdaMART的个性化搜索检索模型

金众威, 刘淑芬, 包铁   

  1. 吉林大学 计算机科学与技术学院, 长春 130012
  • 收稿日期:2015-12-15 出版日期:2016-07-26 发布日期:2016-07-20
  • 通讯作者: 刘淑芬 E-mail:liusf@mail.jlu.edu.cn

Personalized Search and Retrieval Model Based on LambdaMART

JIN Zhongwei, LIU Shufen, BAO Tie   

  1. College of Computer Science and Technology, Jilin University, Changchun 130012, China
  • Received:2015-12-15 Online:2016-07-26 Published:2016-07-20
  • Contact: LIU Shufen E-mail:liusf@mail.jlu.edu.cn

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.

关键词: 决策树, 增强树, 搜索引擎, 个性化搜索

Abstract:

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.

Key words: search engine, decision tree, adaboost decision tree, personalized search

中图分类号: 

  • TP311