挖掘机构别名的Jaccard相似度数据空间转换方法  

Jaccard similarity based data space transform for organization alias mining

作　　者：尚玉玲 曹建军 李红梅 刘艺[1] SHANG Yuling;CAO Jianjun;LI Hongmei;LIU Yi(Commands Institute Information Systems,PLA University of Science and Technology,Nanjing 210007,China;The 63rd Institute,National University of Defense Technology,Nanjing 210007,China)

机构地区：[1]中国人民解放军理工大学指挥信息系统学院,南京210007 [2]国防科技大学第六十三研究所,南京210007

出　　处：《计算机工程与应用》2018年第13期88-92,共5页Computer Engineering and Applications

基　　金：国家自然科学基金面上项目(No.61371196);中国博士后科学基金特别资助项目(No.201003797);中国博士后科学基金面上项目(No.20090461425)

摘　　要：针对同一机构实体对应多个机构名称的问题,提出了一种基于Jaccard相似度数据空间转换的机构别名挖掘方法。根据机构与作者间的隶属关系,建立机构-作者二部图模型;采用Jaccard相似度度量两机构名称所对应作者姓名集合间的相似度;根据机构间的相似度矩阵,将集合型数据转换成数值型数据;通过计算机构名称对应的相似度向量间的余弦相似度,实现了机构别名的有效挖掘。最后用真实数据进行对比实验验证了该方法的优越性。To solve the problem which the same organization entity has few names, a Jaccard Similarity based Data Space Transform for Organization Alias Mining（JS-DST-OAM）method is proposed. Based on the subjection relationship between organizations and authors, organization-author bipartite graph is built; Jaccard similarity is used to measure the similarity of two organization names by their author sets; based on the organization-organization similarity matrix, the transform from set data to numerical data is achieved; cosine similarity of organization name pairs is calculated by their similarity vectors, and it achieves the mining of organization alias. In the end, real data is used to verify its superiority.

关 键 词：实体分辨 机构别名 数据空间转换 Jaccard相似度 余弦相似度 关系数据 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]