基于两种文摘数据的作者相似性探测

Author Similarity Detection Based on Two Abstract Data

[目的/意义]作者相似性探测一直是图书情报领域的热点研究问题之一,现有基于作者合著关系、作者关键词耦合、作者文献耦合等分析方法多假设关键词、标题、引文数据之间相互独立,难以真实准确地反映作者研究内容的相似性。[方法/过程]构建作者的关键词-标题和引文-标题2模矩阵,分别以标题向量表征关键词和引文,再以各关键词和引文的夹角余弦平均值表征作者相似性,并对关键词和引文加权从非对称视角下考察作者的相似性。[结果/结论]实验结果表明,基于加权的关键词-标题和引文-标题数据可以从非对称视角下较为准确地分析作者的相似性。

[Purpose/significance]Research on author similarity has always been one of the hot issues in the field of library and information science.The existing methods based on author co-authoring relationship,author keyword coupling,author document coupling,etc.assume that keywords,titles,and citation data are independent of each other,which is difficult to truly and accurately reflect the similarity of the author’s research content.[Method/process]This paper intends to construct the author’s keyword title and citation-title 2-module matrix,respectively,using the title vector to represent the keywords and citations,and then using the cosine similarity mean of each keyword and citation to represent the author’s similarity,and to investigate the change of author’s similarity under the asymmetric perspective before and after the weighting of keywords and citations.[Result/conclusion]The experiment shows that the weighted keyword title and citation title data can accurately analyze the similarity between authors from an asymmetric perspective.