面向文献搜索系统的用户实时需求发现方法  

Finding method of users' real-time demands for literature search systems

作　　者：徐浩[1] 陈雪[1] 胡晓峰[1] 

机构地区：[1]上海大学计算机工程与科学学院,上海200444

出　　处：《计算机应用》2015年第7期1975-1978,1983,共5页journal of Computer Applications

基　　金：上海市教委创新项目(B.10-0108-14-202)

摘　　要：针对当前文献搜索系统不能理解用户实时需求的问题,提出了一种面向文献搜索系统的用户实时需求发现方法。首先,分析用户浏览、下载等个性化搜索行为;其次,根据用户搜索行为与用户需求的关系构建用户实时需求文档(RD);然后,从用户需求文档中提取用户需求关键词网络;最后,运用随机游走的方法提取出关键词网络的核心节点构成用户需求图。实验结果表明:在模拟用户需求的环境下,提取需求图的方法比K-medoids算法在检索指标F值上平均高2.5%;在用户搜索文献真实情况下,提取需求图的方法比DBSCAN算法在检索指标F值上平均高5.3%,因此,在用户需求比较稳定的文献搜索中,该方法能够获取用户需求从而提升用户体验。Because of the literature search system failing to comprehend users＇ real-time demands, a method to find users＇ real-time demands for literature search systems was proposed. Firstly, this method analyzed the users＇ personalized search behaviors such as browsing and downloading. Secondly, it established users＇ real-time Requirement Documents （RD） based on the relations between users＇ search behaviors and users＇ requirements. And then it extracted keyword network from requirement documents. Finally, it gained users＇ demand graphs which were formed by core nodes extracted from keyword network by means of random walk. The experimental results show that the method by extracting demand graphs increases the F-measure by 2.5%, in the comparison of the K-medoids algorithm on average, under the condition that users＇ demands are emulated in the experiment. And it also increases the F-measure by 5.3%, in the comparison with the DBSCAN （ Density-Based Spatial Clustering of Applications with Noise） algorithm on average, under the condition that users really searches for papers. So, when the method is used in literature search systems where users＇ requirements are stable, it will be able to gain users＇ demands to enhance users＇ search experiences.

关 键 词：用户行为分析 实时需求 文献搜索系统 个性化 关键词网络 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]