搜索引擎用户满意度评估

发布时间：2018-04-15 05:28

本文选题：搜索满意度 + 相关性　；参考：《计算机研究与发展》2017年06期

【摘要】：用户满意度评估一直是互联网搜索领域的研究热点,并具有3方面的挑战:1)传统的搜索性能评估方法大多基于对检索结果相关性的标注,但大多数基于相关性标注的评价指标并非针对互联网搜索环境而设计,其结果与搜索用户主观满意度之间的关系缺乏相应研究;2)大多数已有的工作都基于搜索结果同质化的假设,但随着搜索引擎的发展,异质化的搜索结果元素开始频繁地出现在搜索结果列表中;3)已有的关于搜索满意度评估的工作主要基于用户的点击和查询修改行为开展,但实际搜索中会有大量的用户会话中缺失此类信息.总结了近期为解决这些研究问题开展的实验研究工作:1)构建了用户行为实验系统,分析了结果相关性与用户所感知到的结果效用和满意度之间的关系;2)基于仔细设计的异质化搜索结果页面,定量地分析了垂直搜索结果的质量、展现形式、位置等因素对用户满意度的影响;3)受现有的采用鼠标移动信息进行搜索结果相关性预测的工作启发,提出了在搜索结果页面上抽取用户鼠标移动行为模式并进行满意度评估的方法.实验结果表明:在真实搜索环境下,所提出的方法优于现有的模型.
[Abstract]:User satisfaction evaluation has always been a hot topic in the field of Internet search, and has three challenges: 1) most of the traditional search performance evaluation methods are based on the annotation of the relevance of retrieval results.However, most evaluation indicators based on correlation tagging are not designed for the Internet search environment.The relationship between the results and the subjective satisfaction of the search users is lack of corresponding research. (2) most of the existing work is based on the hypothesis of homogeneity of search results, but with the development of search engines,Heterogeneous search results elements began to appear frequently in the search results list.However, there will be a large number of user sessions in the actual search missing such information.This paper summarizes the recent experimental research work to solve these problems. (1) A user behavior experiment system is constructed.The relationship between the relevance of the results and the perceived utility and satisfaction of the results is analyzed. (2) based on carefully designed heterogeneous search results pages, the quality and presentation of vertical search results are quantitatively analyzed.The influence of location and other factors on user satisfaction is inspired by the existing work of using mouse mobile information to predict the relevance of search results.This paper proposes a method to extract user mouse movement behavior patterns from search results page and evaluate their satisfaction.The experimental results show that the proposed method is superior to the existing model in real search environment.
【作者单位】：清华大学计算机科学与技术系;
【基金】：国家自然科学基金优秀青年科学基金项目(61622208)~~
【分类号】：TP391.3

【相似文献】