基于J-NMF算法的作者类团划分研究  

The Author Similarity Classification Based on J-NMF Algorithm

作　　者：程结晶[1] 李秀霞 Cheng Jiejing

机构地区：[1]扬州大学社会发展学院,江苏扬州225008 [2]曲阜师范大学传媒学院,山东日照276826

出　　处：《情报理论与实践》2020年第11期40-46,共7页Information Studies:Theory & Application

基　　金：国家社会科学基金项目“文献内容分析与引文分析融合的知识挖掘与发现研究”的成果之一,项目编号:16BTQ074。

摘　　要：[目的/意义]为提高作者类团划分的可靠性,针对作者类团划分研究方法缺乏融合作者多源信息、无用信息干扰严重的现状,文章提出一种基于J-NMF算法的作者类团划分方法。[方法/过程]以图书情报学领域154位高产作者为研究对象,利用BICOMB软件生成作者合著矩阵;通过LDA主题提取,构建反映作者文献内容信息的作者主题矩阵;最后,通过J-NMF算法将作者合著信息与作者文献内容信息进行有效融合。[结果/结论]实验表明,相比利用单一作者合著信息和通过融合函数实现的类团划分法,该方法能够有效降低干扰信息,识别相似作者的可靠性更高,作者类团划分能力更强。[Purpose/significance]In order to improve the reliability of author group classification,a method of recognizing author group based on J-NMF algorithm is proposed to overcome the shortage of authors’multi-source information in recognizing author group.[Method/process]Taking 154 high-yielding authors in the field of Library and Information Science as subjects.Firstly,the author co-author matrix is generated by BICOMB software.Then,The author-topic matrix is constructed using LDA Model.Finally,the author’s co-authorship information and the author’s literature content information is effectively fused by J-NMF algorithm.[Result/conclusion]Compared with single co-authorship information and author group recognition method based on fusion function,the proposed method has higher ability of classification of author groups and can identify the similarity between authors more effectively.

关 键 词：J-NMF算法 合著 融合函数 作者类团 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]