面向Web社会网络搜索的人名同一性判断

发布时间：2018-03-29 00:29

本文选题：社会网络　切入点：向量空间模型　出处：《计算机工程与科学》2012年09期

【摘要】：随着人们在互联网上的行为日益丰富,互联网上的社交行为和关系逐渐接近传统的客观世界的社交网络,并能够真实反映出人与人之间在客观世界的真实关系。可以从互联网中通过搜索的方式来构建一个真实客观世界的社会网络。社会网络搜索技术及其方法逐渐成为目前的研究热点,如何对每个Web进行人名同一性判断是社会网络搜索的关键技术。为了从文本中抽取准确的特征并降低向量维度,本文给出了一个基于C-value和逆文档频率IDF的特征向量权值计算方法;实现了基于余弦夹角的相似度计算的算法;通过对文本聚类算法中层次聚类算法和划分聚类算法的研究,给出一种改进的层次聚类算法来实现人名同一性判断。以搜索引擎的人名检索结果进行测试,说明了基于改进的层次聚类算法能有效地提高人名同一性判断的性能。
[Abstract]:As people become more and more active on the Internet, social behaviors and relationships on the Internet are gradually approaching those of the traditional objective world. And can truly reflect the real relationship between people in the objective world. We can build a real and objective social network through the way of searching from the Internet. Social network search technology and its methods gradually become. For the current research focus, How to judge the identity of each Web is the key technology of social network search. In order to extract accurate features from the text and reduce the vector dimension, In this paper, a method of calculating the weight of feature vector based on C-value and inverse document frequency IDF is presented, and the algorithm of similarity calculation based on cosine angle is realized, and the hierarchical clustering algorithm and partition clustering algorithm in text clustering algorithm are studied. An improved hierarchical clustering algorithm is presented to judge the identity of a person's name. The result of search engine's name retrieval shows that the improved hierarchical clustering algorithm can effectively improve the performance of the judgment of the identity of a person's name.
【作者单位】：华中科技大学计算机科学与技术学院;
【基金】：国家自然科学基金资助项目(60873030,61173049)
【分类号】：TP391.1

【参考文献】