基于自组织映射期刊主题可视化组织.pdfVIP

情 报 学 报摇 摇 摇 ISSN 1000-0135 JOURNAL OFTHE CHINA SOCIETY FOR SCIENTIFIC ANDTECHNICAL INFORMATION ISSN 1000 -0135 第30卷第2期183 - 191,2011年2月 Vol.30 No.2,183 - 191 February 2011 試試試試試試試試試試試試試試試試試試試試試試試試試試試試試試試試試試試試試試試試試試試試試試試試 doi:10.3772/ j.issn.1000鄄0135.2011.02.010 基于自组织映射的期刊主题可视化组织1) 安摇 璐摇 李摇 纲 (武汉大学信息管理学院,武汉430072) 摘要摇 摇 本文利用自组织映射(SOM)人工神经网络方法对学术期刊按其主题进行可视化组织。 在修改常见的SOM 显示方式统一距离矩阵(U鄄matrix)的基础上提出增强型U鄄matrix及新的SOM显示方式“属性方差矩阵冶(AV鄄matrix), 构造了“关键属性投影冶方法,以53种有代表性的图书情报类英文期刊为例,将期刊按其主题分为19个类,识别各类 期刊之间的关键差异主题,并分析各类期刊在关键差异主题上的特点。 关键词摇 摇 自组织映射摇 期刊摇 主题摇 可视化 Visual Organization of Journal Subjects with Self鄄Organizing Map An Lu and Li Gang (School of InformationManagement,Wuhan University,Wuhan430072) Abstract摇 摇 This paper aimed to visually organize the academic journals according to their subjects with the Self鄄 OrganizingMap (SOM)technique. AnEnhancedUnifiedDistanceMatrix (U鄄matrix)waspresentedwithsomemodifications on U鄄matrix,a conventional SOM display. A novel SOM display named Attribute Variance Matrix (AV鄄matrix) was also proposed and a method of KeyAttribute Projectionwasconstructed. Fifty鄄threetypical Englishjournalsinthefieldof Library and Information Science (LIS) were selected as experimental samples and categorized into nineteen clusters. The subjects whichcontributedthemosttothedifferencesamongjournalswereidentifiedandthecharacteristicsofjournalclustersinterms of subjectswere analyzed. Keywords摇 摇 self鄄organizing map,journal,subject,visualization 价值的参考。 1摇 引摇 言 学术期刊通常涉及大量的主题,期刊主题研究