基于互信息强度构建标签概念层次结构方法的探究  

Research on the Method of Tag Concept Hierarchy Construction Based on the Strength of Mutual Information

作　　者：江雪琴[1] 张志平[1] 李琳娜[1] 

机构地区：[1]中国科学技术信息研究所,北京100038

出　　处：《情报杂志》2014年第12期165-169,共5页Journal of Intelligence

基　　金："十二五"国家科技支撑计划课题"基于多源信息的电动汽车数据挖掘关键技术研究"(编号2013BAG06B01.)的研究成果之一

摘　　要：大众分类、社会化标注等个性化标签系统对网络资源的组织优势很明显,但存在标签多样和概念结构模糊的缺陷。文章以"豆瓣"读书标签为例,探讨自由标签系统概念上下位关系的识别方法。主要是先对标签进行聚类,将其分成若干个内部结构联系紧密的类簇,然后以互信息强度,具体判别上下位关系,由此构建标签层次结构。实验表明构建完成的概念层次结构图,能够比较准确地识别标签间的概念关系,为用户提供良好的标签导航和浏览机制。Personalized tagging systems, such as folksonomy and social tagging, play an important role in the organization of network re-sources. However, there are the defects of the diversity of the tags used by users and the unclear conceptual structure. In this paper, tags from “Douban” are chosen as the dataset in our experiment. At first, tags are divided into several clusters with closely-linked internal structure. Then, the hyponymy relations of tags are recognized based on the strength of mutual information. The results demonstrate that the concept hierarchical relationship among tags can accurately recognize the hyponymy relations so that it can help the tagging systems to provide users with good tag navigation and browsing mechanisms.

关 键 词：标签 层次关系识别 聚类 互信息强度 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]