组合优化视角下的科技关联识别方法研究  

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

作　　者：黄璐[1] 蔡依洁 陈翔[1] 王长天 HUANG Lu;CAI Yijie;CHEN Xiang;WANG Changtian(School of Management and Economics,Beijing Institute of Technology,Beijing 100081,China;R&D Center,Agricultural Bank of China,Beijing 100005,China)

机构地区：[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

出　　处：《科学学与科学技术管理》2024年第4期118-136,共19页Science of Science and Management of S.& T.

基　　金：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)。

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。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

关 键 词：科技关联 深度学习 网络分析 Node2Vec BERT 

分 类 号：F204[经济管理—国民经济] G301[文化科学] G353.12