基于专利分类号和文本相似度的双元创新识别及测度方法研究  

Research on Identification and Measurement Methods of Ambidextrous Innovation Based on Patent Classification Codes and Text Similarity

作　　者：白全民 王东旭 贾永飞 牛晓霞 王倩 Bai Quanmin;Wang Dongxu;Jia Yongfei;Niu Xiaoxia;Wang Qian(Institute of Science and Technology for Development of Shandong,Qilu University of Technology(Shandong Academy of Sciences),Jinan 250014,China;Think Tank Center,Qilu University of Technology(Shandong Academy of Sciences),Jinan 250014,China)

机构地区：[1]齐鲁工业大学(山东省科学院)山东省科技发展战略研究所,山东济南250014 [2]齐鲁工业大学(山东省科学院)智库中心,山东济南250014

出　　处：《科技管理研究》2025年第1期10-17,共8页Science and Technology Management Research

基　　金：山东省社会科学规划研究项目“基于企业创新响应的山东省未来产业培育发展路径研究”(24CGLJ09);山东省“泰山学者”工程专项经费项目“科技创新支撑黄河流域生态保护和高质量发展的策略研究”(tsqn202312232)。

摘　　要：学术界关于如何准确、有效地识别与测度双元创新尚未达成一致定论。通过明晰双元创新概念的内涵与外延,认为双元创新理论关注的是企业相对自身既有技术能力的创新行为,这是进行识别与测度方法设计的逻辑起点,因此在传统专利方法的基础上,引入技术内容相关性维度,对探索式创新进行解构,构建识别与测度企业双元创新的新方法。利用中国人工智能领域188家上市企业的98534件发明专利,借助Python编程工具,以及余弦相似度算法,对设计的识别与测度方法进行验证。结果显示:大部分人工智能企业仍以利用式创新为主,在探索式创新中也更倾向于开展探索性较弱的创新活动;人工智能技术作为渗透能力较强的通用性技术,大幅拓展企业探索式创新空间;双元创新是一个相对概念,利用式创新和探索式创新间的均衡程度决定测度得分的高低。基于此,从政府、企业、社会机构3个方面提出管理启示。The accurate identification and measurement of ambidextrous innovation(exploitative innovation and exploratory innovation)remain unresolved in academic research.This paper clarifies the connotation and extension of the ambidextrous innovation concept,emphasizing that ambidextrous innovation focuses on enterprises'innovative behaviors relative to their existing technological capabilities.Building on traditional patent analysis methods,it introduces the dimension of technical content relevance to deconstruct exploratory innovation,thereby constructing a new method for identifying and measuring ambidextrous innovation in enterprises.Using 98534 invention patent applications from 188 listed companies in the field of artificial intelligence(AI)in China,it tests and verifies the proposed method with the aid of Python programming tools and cosine similarity algorithms.The empirical results reveal that:(1)Exploitative innovation dominates.most AI enterprises still primarily rely on exploitative innovation,with exploratory innovation accounting for only 14.10%of total patents.Among exploratory innovations,typeⅢ(the most exploratory)has the lowest proportion(2.12%),indicating that enterprises are more inclined to engage in less risky and less exploratory innovation activities.(2)AI technology broadens the horizons of exploratory innovation.As a universal technology with strong penetration capabilities,AI significantly broadens the exploratory innovation space for enterprises,empowering them to pursue new technological trajectories.(3)Ambidextrous innovation is a relative concept.The balance between exploitative and exploratory innovation determines the measurement scores,highlighting the importance of maintaining a strategic equilibrium between the two types of innovation.Based on these findings,the paper proposes management implications from the perspectives of government,enterprises,and social institutions.For policymakers,it is crucial to consider the dual dimensions of innovation when designing support policies.Enter

关 键 词：利用式创新 探索式创新 专利分析 识别方法 创新测度 人工智能 

分 类 号：F273.1[经济管理—企业管理] F272.3[经济管理—国民经济] F224[文化科学] G301