摘要
[目的/意义]挖掘海量文章中的研究主题,梳理研究主题的演化脉络和关联,预测主题前沿热点,增强演化结果的科学性和生动性。[方法/过程]文章提出时序影响因子的概念作为关键词提取中的重要特征,采用时间窗口的方法,利用主题模型挖掘与主题识别,并进行可视化分析,通过对深度学习领域的时间序列模型的应用,达到预测主题流行度的目的。[结果/结论]实验验证了融合时序特征的关键词提取,可以提升主题模型的效果,通过可视化方式既能观察到整体的主题流行度变化趋势,也能对各个时间段的主题内容进行演化分析,观察其分裂和合并的趋势。
[Purpose/significance]Excavating the research topics in a large number of articles,sorting out the evolution context and correlation of the research topics,predicting the frontier hot spots of the topics can be helpful to enhance the scientificity and vividness of the evolution results.[Method/precess]This paper puts forward the concept of time series influence factor as an important feature in keyword extraction,uses the method of time window to mine and identify topics by using topic model,and makes visual analysis.By applying time series model in the field of deep learning,the purpose of predicting topic popularity is achieved.[Result/concluson]It is verified that the keyword extraction integrating temporal features can improve the effect of the topic model.Through visualization,not only the change trend of the overall topic popularity can be observed,but also the evolution of the topic content in each time period can be analyzed,its splitting and merging trend can be also explored.
作者
李树青
朱军涛
王婉
LI Shuqing;ZHU Juntao;WANG Wan(College of Information Engineering,Nanjing University of Finance and Economics,Nanjing 210023;Hunan Institute of Scientific and Technical Information,Changsha 410001)
出处
《科技情报研究》
CSSCI
2023年第2期57-77,共21页
Scientific Information Research
基金
国家社会科学基金项目“学术虚拟社区知识交流效率研究”(编号:17BTQ028)
江苏省研究生科研与实践创新计划项目“基于时序关键词特征的学术主题热点及演化趋势研究”(编号:WWXW21001)。
关键词
主题演化
关键词抽取
可视化分析
timing keywords
theme evolution
keyword extraction
visual analysis
