情报科学 ›› 2025, Vol. 43 ›› Issue (9): 182-190.

• 博士论坛 • 上一篇    下一篇

融合BERTopic和DTM模型的技术主题识别及演化研究 ——以美国生物技术领域为例

  

  • 出版日期:2025-09-05 发布日期:2025-12-12

  • Online:2025-09-05 Published:2025-12-12

摘要: 【目的/意义】技术主题识别与演化是追踪技术创新趋势的有效手段,使用定量方法识别政府资助企业产品 开发研究项目的技术主题,能够为政府引导企业技术创新提供决策支撑,也为生物技术领域科研人员选题提供理 论和实践依据。【方法/过程】探究BERTopic静态主题模型融合DTM动态主题模型的技术主题识别和演化分析框 架。首先,构建BERTopic静态主题模型识别技术主题,通过可视化方法解析技术主题相似度;其次,在语料库中构 建时间窗口,采用DTM动态主题模型解析不同时间窗口下的技术主题演化趋势;最后,结合统计分析不同技术主 题的政府关注度及未来研究热度。【结果/结论】采用所构建的研究框架识别出美国政府资助生物技术领域的十大 技术主题,并对其技术演化长期趋势进行分析。【创新/局限】创新性的结合BERTopic模型对政府资助企业产品开 发研究项目数据进行技术主题识别,但研究采用的数据单一,未来拟从多源多模态数据视角优化研究。

Abstract: 【Purpose/significance】Identifying and tracking the evolution of technical topics is an effective means of monitoring techno⁃ logical innovation trends. Utilizing quantitative methods to identify the technical themes of government-funded enterprise product de⁃ velopment research projects can provide decision-making support for guiding corporate technological innovation. It also offers theoreti⁃ cal and practical foundations for researchers in the biotechnology field to select research topics.【Method/process】This study explores a framework for technology theme identification and evolution analysis by integrating the BERTopic static topic model with the DTM dynamic topic model. First, the BERTopic static topic model is constructed to identify technological themes, using visualization meth⁃ ods to analyze the similarity of these themes. Next, a time window is established within the corpus to apply the DTM dynamic topic model, which analyzes the evolutionary trends of technological themes across different time windows. Finally, a statistical analysis is conducted to examine the government's attention to different technological themes and their future research intensity【. Results/conclu⁃ sion】Using the proposed research framework, the top ten technological themes in the biotechnology field funded by the U.S. govern⁃ ment were identified, and a long-term trend analysis of their technological evolution was conducted【. Innovation/limitation】The study innovatively integrates the BERTopic model to identify technology themes in government-funded enterprise product development re⁃ search project data. The study’s limitation lies in its reliance on a single data source, and future research aims to optimize the frame⁃ work from multi-source and multi-modal data perspectives.