大模型辅助的研发合作机会识别与可交互图谱构建  

R&D Cooperation Opportunities Identification and Interactive Graph Construction Assisted by Large Language Model

作　　者：郝世博[1] 史东昊 朱学芳[2] 孙明汉 宦昭润 Hao Shibo;Shi Donghao;Zhu Xuefang;Sun Minghan;Huan Zhaorun(Nanjing University of Science and Technology,Nanjing 210094;Nanjing University,Nanjing 210023)

机构地区：[1]南京理工大学知识产权学院,南京210094 [2]南京大学信息管理学院,南京210023

出　　处：《情报杂志》2025年第4期173-180,189,共9页Journal of Intelligence

基　　金：国家社会科学基金项目“5G环境下中国智慧知识服务体系构建研究”(编号:22BTQ017)研究成果。

摘　　要：[研究目的]为提高科技成果转化效率和质量、优化创新资源配置,实现技术供需精准对接和挖掘潜在研发合作机会。[研究方法]首先通过大语言模型辅助智能抽取得到关键技术信息,利用不同模型获得文本嵌入向量表征并生成相似度矩阵;其次使用最优模型编码的专利相似度数据生成创新主体相似度,并构建满足多元化场景挖掘需求的研发合作机会可交互网络图谱;最后以激光增材制造领域为例验证了该文所提方法的可行性。[研究结果/结论]案例研究表明,所提出的研发合作机会识别与可交互网络图谱构建方法能够为企业技术创新及高校院所专利转化运用提供更为精准的决策支撑。[Research purpose]To improve the efficiency and quality of the transformation of scientific and technological achievements,optimize the allocation of innovative resources,achieve precise matching of technology supply and demand,and explore potential opportunities for research and development cooperation.[Research method]Firstly,utilize aided intelligent large language model to extract key technical information,and then obtain text embedding vector representations and generate the similarity matrix by means of different models.Next,employ the patent similarity data encoded by the optimal model to generate the innovative subject similarity,and construct an interactive network graph of R&D cooperation opportunity meeting the demands of diversified scenario mining.Finally,verify the feasibility of the proposed method taking the laser additive manufacturing as an example.[Research result/conclusion]The case study indicates that the proposed method for identifying the R&D cooperation opportunity and constructing the interactive network graph can provide more accurate decision support for enterprises'technological innovation and the transformation of patents in institutions of higher learning and research institutes.

关 键 词：专利文本嵌入 大语言模型 研发合作机会识别 可交互网络图谱 激光增材制造 

分 类 号：G350[文化科学—情报学]