以分类主题抽取为视角的学科主题挖掘——基于LDA模型的国外知识流研究结构探讨

Subject Topic Mining from the Perspective of Classified Topic Extraction

[目的]为了能够更为全面地探索和揭示研究领域的知识结构和热点主题,文章提出基于分类视角的LDA主题抽取方法。[方法]以国外知识流领域为研究对象,根据研究方向将知识流的相关文献分为5类。利用LDA主题模型分别对分类后的文献集进行主题抽取,筛选得到不同研究方向下的11个热点主题,并深入分析不同研究方向下热点主题所揭示的知识点。[结果]实验结果表明,基于分类视角的LDA主题抽取方法能够较为全面和细致地挖掘研究领域的学科主题和研究热点。[局限]所提的方法未能与其他主题挖掘方法进行对比,研究结果也未与现有文献中分析出的知识流领域研究热点进行对照。

[ Purpose ] In order to help explore and reveal the knowledge structure and hot topics of research field to a full ex-tent, this paper proposes LDA （Latent Dirichlet Allocation） topic extraction method based on the perspective of classification.[ Method] The paper takes foreign knowledge flow （KF） field as the research object, and classifies the relevant literature of KF in-to 5 categories according to research areas. The paper uses LDA topic model to extract the topic of classified literature collections andselects 11 hot topics for different research areas, in which knowledge hotspots are thoroughly analyzed to reveal the structure map ofKF filed. [ Result] Experimental results indicate that the method proposed in the paper can detect discipline theme and researchhotspots in a more comprehensive and in-depth way. [ Limitations The paper does not compare the proposed method with othertopic mining methods and contrast hotspots of KF field in the result with those in the existing literature.