低频查询的用户行为分析和类别研究

发布时间：2018-06-09 21:02

本文选题：低频查询 + 用户行为　；参考：《计算机研究与发展》2012年11期

【摘要】：低频查询是用户提交查询频次非常低的查询.它们占了搜索引擎独立查询的很大比例且对用户体验影响巨大,但由于数据稀疏性,现有的搜索引擎用户行为分析及相关研究中对低频查询涉及很少.结合前人的相关工作,使用商业搜索引擎的大规模用户日志,在会话级别上进行低频查询的用户行为分析以及类别研究.基于目标查询行为、后续相关查询行为、整体会话行为3个方面的12个特征进行了低频查询的用户行为特征分析,首次提出了低频查询类别分析框架,并进一步使用改进的AdaBoost算法对低频查询会话进行分类.实验对2000个随机的低频查询会话样例进行分类,AUC值达到了83%以上.低频查询的用户行为分析和类别研究,将为搜索引擎用户行为分析等网络检索研究提供重要基础.
[Abstract]:Low frequency query is a very low frequency query submitted by the user. They account for a large proportion of search engine independent queries and have a great impact on the user experience. However due to the sparsity of data the existing search engine user behavior analysis and related research on low-frequency queries are rarely involved. Combined with the related work of predecessors, using the large-scale user log of commercial search engine, the user behavior analysis and category research of low-frequency query at session level are carried out. Based on 12 features of target query behavior, subsequent related query behavior and overall conversation behavior, the user behavior characteristics of low frequency query are analyzed, and a low frequency query category analysis framework is proposed for the first time. Furthermore, the improved AdaBoost algorithm is used to classify low frequency query sessions. The AUC value of 2000 random low frequency conversation samples is over 83%. User behavior analysis and classification research of low frequency query will provide an important basis for search engine user behavior analysis and other network retrieval research.
【作者单位】：智能技术与系统国家重点实验室;清华大学计算机科学与技术系;
【基金】：国家“八六三”高技术研究发展计划基金项目(2011AA01A205) 国家自然科学基金项目(60903107,61073071) 高等学校博士学科点专项科研基金项目(20090002120005)
【分类号】：TP391.3

【参考文献】