科学学与科学技术管理

科学学与科学技术管理 ›› 2024, Vol. 45 ›› Issue (04): 118-136.

• 创新战略与管理 • 上一篇    下一篇

组合优化视角下的科技关联识别方法研究

  

  1. 1. 北京理工大学 管理与经济学院，北京 100081；2. 中国农业银行研发中心，北京 100005
  • 收稿日期:2023-03-01 出版日期:2024-04-10 发布日期:2024-04-28
  • 通讯作者: 陈翔，bjchenxiang@hotmail.com
  • 作者简介:黄璐（1984— ），女，汉族，江西抚州人，北京理工大学管理与经济学院教授，博士，研究方向：科技创新管理；蔡依洁 （2000— ），女，汉族，北京人，北京理工大学管理与经济学院硕士，研究方向：科技创新管理；陈翔（1976— ）男，汉族，江西赣 州人，北京理工大学管理与经济学院教授，博士，研究方向：数据挖掘；王长天（1998— ），男，汉族，江西赣州人，中国农业银行 研发中心，硕士，研究方向：复杂网络。
  • 基金资助:
    国家自然科学基金面上项目（72274013；72371026）；北京市社会科学基金决策咨询项目（23JCB021）

Looking for the Best Allocation of Scientific and Technological Resources: A Perspective of Combinatorial Optimization

  1. 1. School of Management and Economics, Beijing Institute of Technology, Beijing 100081, China; 2. R&D Center, Agricultural Bank of China, Beijing 100005, China
  • Received:2023-03-01 Online:2024-04-10 Published:2024-04-28

PDF (PC)

16

摘要： 以论文数据表示科学，专利数据表示技术，构建了一套基于深度学习的科学—技术关联识别方法体系。首先，利 用 Node2Vec 和 BERT 模型获得论文和专利关键词的知识结构表示和文本语义表示，构建科学网络和技术网络；之后，运 用 Fast Unfolding 社区发现算法和 Z-Score 指标精准识别科学主题和技术主题；最后，构建科学—技术主题完全二分图， 将科学—技术主题关联识别问题转化为二分图匹配问题，利用Kuhn-Munkres算法求解最优科技关联匹配。基于2010— 2021年“自然语言处理”领域的论文与专利数据开展实证分析，验证研究方法的有效性。

关键词: 科技关联, 深度学习, 网络分析, Node2Vec, BERT

Abstract: As the integration of science and technology accelerates in the present era, the characteristics of their mutual interaction, combination, penetration, and transformation have become increasingly pronounced. In-depth exploration of the knowledge linkages between science and technology (S&T) is an essential prerequisite for accurately understanding the S&T innovation laws, promoting the transformation of scientific outcomes, and optimizing S&T innovation policies. However, there is a dearth of research that effectively captures the information from both the knowledge structure and textual semantics of science and technology, let alone deeply explores the linkage from the perspective of achieving optimal matching between science and technology topics. A novel deep learning-based methodology is proposed to investigate S&T linkages, where papers and patents are applied to represent science and technology. Specifically, science and technology networks are constructed based on Node2Vec and BERT. Then, science and technology topics are identified based on the Fast Unfolding algorithm and Z-Score index. Finally, a science-technology bipartite graph is constructed, the S&T topic linkages identification task is successfully transferred into a bipartite matching problem, and the maximum-weight matching is identified using a Kuhn-Munkres bipartite algorithm. Based on this, an empirical analysis is carried out using paper and patent data from the field of "Natural Language Processing" from 2010 to 2021. In validation, the proposed method is compared with four network construction methods in terms of topic identification, and its effectiveness is further validated against keywords linkage method and two semantic similarity methods in terms of topic similarity measurement. The results reveal that in the periods 2010-2013, 2014-2017, and 2018-2021, 82, 51, and 91 science-technology topic pairs are identified respectively. From 2010 to 2013, interactions in the NLP field began to increase, but the depth of linkage was superficial, mainly focusing on exploring ways to improve the performance of existing models and systems. From 2014 to 2017, although the frequency of science and technology interactions slightly reduced, a more profound fusion of science and technology had been achieved. It is worth noting that many interactions in this period between S&T appear in discovering the role of existing scientific theories in the new technology application scenarios. From 2018 to 2021, innovation activities in the NLP field entered a vibrant phase, with both the intensity and depth of S&T linkage significantly increasing, and research and applications of multi-modality data became a new trend. The primary theoretical contributions are as follows. First, the comprehensive application of Node2Vec and BERT deep learning methods achieved the effective integration of knowledge structure and textual semantic information, deepening the application of deep learning techniques and semantic analysis methods in S&T linkage research. Second, innovatively integrating network analysis methods, constructing the“topic coupling strength”indicator to capture the rich network structure information in scientific and technological knowledge systems, offering valuable additions to the study of S&T interactions and collaborative innovation pathways. Third, converting the science-technology topic linkage identification problem into a bipartite graph matching problem, realizing the combinatorial optimization of the S&T knowledge systems, providing a fresh perspective for S&T linkage analysis. Fourth, enriching the research in fields related to industry-academia-research collaborative innovation and innovation ecosystem governance, providing essential theoretical support for promoting the transformation of basic research results and driving the deep integration of innovation and industrial chains.

Key words: science-technology linkages, deep learning, network analysis, Node2Vec, BERT

中图分类号: