情报科学 ›› 2021, Vol. 39 ›› Issue (5): 146-155.

• 业务研究 • 上一篇    下一篇

基于主题时态关联的科学领域研究演化识别

  

  • 出版日期:2021-05-01 发布日期:2021-05-12

  • Online:2021-05-01 Published:2021-05-12

摘要:

【目的/意义】从主题时态的角度,构建主题时态关联的科学领域研究演化的识别方法,为研究人员精准把
握领域发展轨迹和科研创新提供参考。【方法/过程】提出基于主题时态关联的领域研究演化识别三阶段模型。首
先利用TF-IDF模型提取领域文献数据集的特征词;再以特征词作为关联规则算法的挖掘条件,生成并过滤主题时
态;最后,依据主题时态关联关系和强度值,构建主题时序演化路径可视化图谱进行领域研究演化识别分析。【结
果/结论】以在线评论有用性领域外文科技文献为实证研究对象进行主题时态关联演化分析,识别出影响因素有用
性、消费者意愿及应用、研究理论技术和有用性排序等主要领域研究的发展演化过程,验证了方法模型的有效性。
【创新/局限】清晰地描绘了主题在时间序列上形成、发展和关注强度的动态演化过程,使得主题演进脉络和关联更
加具有可读性,但样本数据未实现全覆盖,对领域研究的整体态势分析不够全面。

Abstract:

【Purpose / significance】Form the perspective of topic tense, this paper constructs an identification method of scientific re⁃
search evolution based on topic tense, which can provide reference for researchers to accurately grasp the development path of the
field and scientific research innovation.【Method / process】A three-stage model of domain research evolution recognition based on top⁃ic temporal association is proposed. Firstly, the TF-IDF model is used to extract the feature words of the domain literature data set;secondly, the feature words are used as the mining conditions of the association rule algorithm to generate and filter the topic tense; fi⁃nally, according to the topic temporal association relationship and strength value, the visual Atlas of the topic temporal evolution path is constructed for the domain research evolution identification and analysis.【Result/conclusion】Taking the foreign scientific and Tech⁃nological Literature in the field of online review usefulness as the empirical research object, the topic temporal correlation evolution analysis was carried out to identify the development and evolution process of the main research fields, such as the usefulness of influ⁃encing factors, consumer willingness and application, research theory and technology, and usefulness ranking, which verified the effec⁃tiveness of the method model.【Innovation / limitation】It clearly describes the dynamic evolution process of topic formation, develop⁃ment and attention intensity in time series, which makes topic evolution context and relevance more readable. However, the sample da⁃ta does not achieve full coverage, and the overall situation analysis of field research is not comprehensive enough.