基于LDA模型的欧盟框架计划近10年农业科研项目研究主题发现

Topic Discovery of Agricultural Scientific Research Projects of EU's Research Framework Program Based on LDA Model in Recent Decade

当前科研项目研究主题发现的方法主要是基于文献计量学,即通过统计项目的外部特征进行分析,但并未涉及项目研究内容。科研项目数据不仅是集技术情报、经济情报、商业情报于一体的知识载体,同时还具有科学性、具体性、探索性、创新性等特点。若能使用有效的方法对科研项目数据内容进行挖掘,将对科研工作者的研究创新以及国家科研政策制定产生积极影响。探讨了基础传统文献计量学方法的科研项目分析的局限性及存在的问题,对比分析了Latent Dirichlet Allocation(LDA)模型在主题发现领域的优势,提出了基于LDA模型的科研项目主题发现方法,并通过对欧盟框架计划近10年的农业科研项目数据进行分析,验证LDA模型在科研项目数据分析中的有效性。基于欧盟农业科研项目研究主题的分析结果发现,基于LDA模型在科研项目主题发现上能够充分挖掘科研项目的主题信息,是一种行之有效的科研项目分析方法。

The current topic analysis methods of scientific research projects are mainly based on the bibliometrics, namely analyzing through the external characteristics of statistical projects, but the content is not involved in the analysis. The data of scientific research projects is not only a set of technical information, economic information, business intelligence, but also has many characteristics such as scientific, concrete, exploratory and innovative. So, if we can use effective data mining method for the contents of scientific research projects, it will have a positive impact on researchers and policy making of national scientific research. The authors discussed the limitations and existing problems of the scientific research projects analysis of the traditional research methods ofbibliometrics, analyzed the advantages of the LDA model in the topic discovery field, put forward the topic discovery method of scientific research projects based on the LDA model, and then analyzed nearly decade data of agricultural scientific research projects of European Framework Program to verify the effectiveness of LDA model in the scientific research projects data analysis. Based on the research topics analysis of EU＇s agricultural research projects, it was found that the LDA model could fully excavate the thematic information of scientific research projects in the topic discovery of scientific research projects, which was an effective method of scientific research projects analysis.