面向用户行为的页面质量评估

Web Page Quality Estimation with User Behavior

页面质量评估在搜索引擎系统中具有极其关键的作用,传统的方法是基于页面链接关系进行页面质量评估。但由于当前Web环境的复杂性,传统方法已经难以适应当前的Web环境,近年来,用户行为被用来弥补完全依赖链接关系方法的不足。用户行为可以分为两类:浏览行为和搜索行为。利用浏览行为构造了用户浏览图;提出了一种利用用户搜索行为的新方法,此方法构造了用户搜索图;合并用户浏览图和用户搜索图得到用户浏览搜索图。实验表明用户浏览搜索图的性能比较接近用户浏览图的性能,并超过全网的性能,同时用户浏览搜索图能够评价的页面数要大于用户浏览图。

Page quality estimation in the search engine has a crucial role, and the traditional method is based on hyperlink structure analysis. Because of the complexity of the current Web environment, the traditional method cannot work well. To solve this, user behavior is paid much attention these years. User behavior can be divided into two types： Browsing behavior and searching behavior. With analysis into browsing behavior, user browsing graph is constructed. A new approach to use searching behavior is proposed, and user searching graph is constructed through this method. With the combination of these two graphs, user browsing-searching graph is constructed. Experimental results show that the performance of user browsing-searching graph is close to user browsing graph, and exceeds the whole Web graph. On the other side, the number of page that the user browsing-searching graph can evaluate is more than user searching graph.