基于专利实体语义表示的技术主题演化路径识别  

Identification of Technological Topic Evolution Paths Based on Patent Entity Semantic Representation

作　　者：张金柱[1] 张毅 Zhang Jinzhu;Zhang Yi(Department of Information Management,School of Economics and Management,Nanjing University of Science and Technology,Nanjing 210094)

机构地区：[1]南京理工大学经济管理学院信息管理系,南京210094

出　　处：《情报杂志》2024年第11期117-128,共12页Journal of Intelligence

基　　金：国家自然科学基金面上项目“基于专利多模态内容和交易数据的互补技术识别与挖掘研究”(编号:72374103);国家自然科学基金面上项目“基于表示学习的专利信息语义融合与深度挖掘研究”(编号:71974095);江苏省研究生科研与实践创新项目(编号:KYCX24-0793)研究成果。

摘　　要：[研究目的]从专利实体抽取和语义表示角度,识别语义相同但表达方式不同的专利实体,更准确地发现技术主题演化路径,更好地辅助科技创新和管理决策。[研究方法]提出一种基于专利实体语义表示的技术主题演化路径识别方法。首先,构建BERT-BiLSTM-CRF模型自动抽取专利实体,利用表示学习方法研究专利实体的语义向量表示。其次,基于K-means算法对实体向量进行聚类,识别技术主题。最后,基于实体语义相似度,识别语义相同但表达不同的专利实体,进而基于相同实体数量设计知识流入和知识流出指标,根据主题之间的知识流入和流出比例共同识别分裂、发展、融合等演化关系,构建技术主题演化路径。[研究结论]实证研究表明,该方法能有效识别语义相同但表达不同的专利实体,进而更加准确地识别主题间演化关系,构建技术主题演化路径。[Research purpose]From the perspective of patent entity extraction and semantic representation,identifying patent entities with the same semantic meaning but different expression methods can more accurately discover the evolution path of technological themes and better assist technological innovation and management decision-making.[Research method]A technology topic evolution path identification method based on patent entity semantic representation is proposed.Firstly,the BERT-BiLSTM-CRF model is constructed to automatically extract patent entities,and the semantic representation of technical entities is realized by representation learning method.Secondly,relevant technical topics are identified based on K-means clustering algorithm.Finally,based on entity semantic similarity,patent entities with the same semantics but different expressions are identified.The knowledge flow index is designed to identify the relationships among topics,such as division,development,and fusion,according to the knowledge flow ratio among topics,and to construct the evolution path of topics.[Research conclusion]The empirical study in the field of artificial intelligence shows that the proposed method can identify the technical entities representing the same meaning from the semantic,and can construct the technological topic evolution path more accurately,which proves the effectiveness and superiority of the proposed method.

关 键 词：专利实体 实体抽取 实体语义表示 BERT-BiLSTM-CRF模型 主题演化 

分 类 号：G350.7[文化科学—情报学]