一种基于网页块特征的多级网页聚类方法  

A multi-level page clustering method based on page segmentation

作　　者：范意兴 郭岩[1] 李希鹏[1,2] 赵岭[1] 刘悦[1] 俞晓明[1] 程学旗[1] 

机构地区：[1]中国科学院计算技术研究所,北京100190 [2]中国科学院大学,北京100190

出　　处：《山东大学学报（理学版）》2015年第7期1-8,共8页Journal of Shandong University(Natural Science)

基　　金：国家重点基础研究发展计划("973"计划)项目(2012CB316303);国家高技术研究发展计划("863"计划)项目(2012AA011003);国家科技支撑计划项目(2012BAH39B02);国家自然科学基金资助项目(61232010;61202058)

摘　　要：利用网页的结构特征,提出一种多级网页聚类方法。该方法首先对网页进行分块,然后使用网页的块特征对网页进行聚类。在聚类过程中,通过调整阈值,能够提供三级聚类:同站点网页聚类、同站点同结构网页聚类、同站点同结构同模板网页聚类。与已有的网页聚类方法相比较,该方法能够提供多级聚类结果,满足不同的聚类需求,而且在聚类的准确率和效率方面有本质上的提高。A multi-level page clustering method based on page segmentation was proposed. In this method, pages were divided into several blocks, and then clustered by using the block feature. By adjusting the threshold of similarity between pages, three-level clustering was obtained： the first level is pages from the same website, the second level is pages from the same website with the same structures, and the last level is pages produced with the same template from the same website. Compared with traditional methods, this method not only could provide multi-level clustering, but also can cluster pages effectively.

关 键 词：网页分块 网页聚类 DOM 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]