摘要
[目的/意义]现有基于合作或引证建立的学术共同体展现了更显性的学术关联,但不能直观地揭示出学术共同体所共有的特征,同时不可避免增加了人情因素带来的偏私倾向。[方法/过程]以知网文献摘要数据为研究对象,文章利用LDA和Word2vec混合模型挖掘得到每篇文献的主题,主题包含主题词及其扩展词。并以此作为主题与文献作者关系的依据,构建学者-主题二模网络,通过对二模网络以及映射的一模学者网络进行可视化,直观地反映了领域内学者就研究方向的聚集情况。[结果/结论]LDA和Word2vec混合模型能够深入挖掘文献主题,而利用二模能够展现二元的主体,通过上述方法,能够找到在现实中或许没有发生合作、但具有潜在重合研究主题倾向的学者群体。以国内图情领域为例,识别其核心学术共同体。"学者-主题"的二模网络中纳入了学者隶属群体的信息,不仅从全局视域归纳出领域内由词语元素构成的具体主题,而且利用向量距离计算得到的各个主题的扩展词语集,能进一步解释学者共同体所隶属群体的深化特征,能够有效降低人情因素,为同行评议提供支持。
[Purpose/significance]The existing academic community based on cooperation or citation reveals more explicit academic association,but at the same time,it fails to reveal the common characteristics of the academic community,as well as to avoid the tendency of partiality caused by human factors.[Method/process]Taking the data of the literature abstracts in CNKI as the research object,this study uses the LDA and Word2 Vec model to excavate the theme including words and extended theme words of each document,which are the basis for the relationship between the subjects and authors of the literature,and constructs the scholar-theme two-mode network.Using the cross-product method,it obtains the one-mode scholar network transformed from the two-mode network.The mapping of the one-mode scholar network intuitively reflects gathering situation of scholars’research directions in the selected field.[Result/conclusion]The LDA and Word2 Vec model can dig deep into the themes of literature while the use of two-mode network can show the dual subject.Through the above methods,the scholars with coincident topics could be found who in reality may not have cooperation.Specifically,for the field of library and information science,the paper identified the core academic community.The"scholars-themes"two-mode network,which includes the information of the scholars’affiliation groups,not only summarizes specific topics composed of keywords from the global view,but also further explains the deepening characteristics of academic community via calculating distance of vectors of extended word sets.To some extent,the method can increase objectivity and provide support to peer review.
出处
《情报理论与实践》
CSSCI
北大核心
2020年第5期170-176,共7页
Information Studies:Theory & Application
基金
华东师范大学2019年度“幸福之花”基金先导项目(人文社会科学)“大数据视阈下基于学术共同体的人文社科学术评价与促进研究”的成果之一,项目编号:2019ECNU-XFZH016。
