技术特征相似性视角下的新兴技术主题识别  

Emerging Technology Topic Recognition from the Perspective of Technology Feature Similarity

作　　者：宋博文 栾春娟 梁丹妮 SONG Bo-wen;LUAN Chun-juan;LIANG Dan-ni(Institute of Humanities&Social Sciences,Dalian University of Technology,Dalian 116024;School of Intellectual Property,Dalian University of Technology,Dalian 116024)

机构地区：[1]大连理工大学人文与社会科学学部,辽宁大连116024 [2]大连理工大学知识产权学院,辽宁大连116024

出　　处：《软科学》2023年第12期80-85,108,共7页Soft Science

基　　金：国家自然科学基金项目(71774020);辽宁省经济发展研究课题(2018lslktqn-016)。

摘　　要：结合专利文本特点提出具有针对性的特征工程思路,整合BERT训练模型与语义分析技术进行新兴技术主题识别模型的构建。结果显示,基于大语言模型的分析思路能够提升识别结果的准确性与解释性。实证分析部分以纳米技术为例,验证了模型的有效性和准确性。Emerging technology is a general term for a class of emerging technologies with the same novelty features.It is of great significance to optimize national science and technology strategic deployment and enhance international competitive-ness.Given the shortcomings of the current research on emerging technology topic recognition,this study proposes an emer-ging technology topic recognition model based on deep learning and semantic analysis from the perspective of technical fea-ture similarity.The first step is to extract the novelty information in the patent data as the technical feature set;the second step,the BERT pre-training model,is used to train the existing technical feature set,and the new technical features are vectorized;the third step is to cluster technical features and extract the core semantic structure of clustering by semantic a-nalysis;and combined with expert opinions.The research results expand the research ideas and methods of emerging tech-nology topic recognition.

关 键 词：深度学习 新兴技术主题 技术特征 主题识别 纳米技术 

分 类 号：G305[文化科学]