WRM一种基于单词相关度的文档聚类新方法.pdfVIP

WRM ：一种基于单词相关度的文档聚类新方法 伍赛* * * * + + 杨冬青 韩近强 张铭 王文清 冯英 * （北京大学信息与科学技术学院 北京 100871） + （北京大学图书馆中国高等教育文献保障系统管理中心 北京 100871 ） （wsai@ ） 摘要 目前大多数的搜索引擎如 Google、百度等，查询的结果都是按照重要度排序然后分页地显示给用户。但 是有时候这样显示并不能很好地服务于用户，用户经常要浏览了很多页面才找到自己所需要的内容。如果将返 回的结果再进行分类，就可以很好的解决这一问题。不同于传统的向量空间模型的方法，本文提出了一种基于 单词相关度的聚类方法。实验的结果表明该方法具有较高的准确性和很高的效率。 关键字 文档聚类，单词相关度，单词向量空间模型 WVM ，向量空间模型VSM ，TF/IDF，聚类引擎 中图法分类号 TP311 WRM: A Novel Document Clustering Method Based on Word Relation * * * * + + Wu Sai Yang Dong-Qing Han Jin-Qiang Zhang Ming Wang Wen-Qing Feng Ying * ( School of Electronics Engineering and Computer Science, Peking University, Beijing, China, 100871) + ( Administrative Center for China Academic Library Information System Room 607, Peking University Library Beijing, China, 100871) Abstract The most popular search engines, such as Google and Baidu, answer users’ queries as lists of ranked results according to importance. But in some cases the most “important” is not the most useful for the user. A user has to look through several pages to get what he wants. Trying to classify the results is a good idea to solve this problem. In this paper, we propose a novel clustering method based on the word relation WRM, which is different from the traditional VSM method. Experiment results show that our method WRM is not only very effective but also efficient. Keywords Document Clustering, Word Relation, Word Vector Model (WVM), Vector Space Model (VSM) , TF/IDF, Clustering Engine