基于CLIP-LDAGV多模态信息融合的颠覆性技术主题识别研究——以新能源领域为例  

Research on Disruptive Technology Topic Recognition Based on CLIP-LDAGV Multimodal Information Fusion-A Case Study of the New Energy Field

作　　者：吕鲲 张未旭 靖继鹏[4] Lyu Kun;Zhang Weixu;Jing Jipeng(Business School,Ningbo University,Ningbo 315211;Merchants’Guild Economics and Cultural Intelligent Computing Laboratory of Ningbo University,Ningbo 315211;College of Computer,National University of Defense Technology,Changsha 410073;School of Business and Management,Jilin University,Changchun 130012)

机构地区：[1]宁波大学商学院,宁波315211 [2]宁波大学“商帮经济与文化”智能计算实验室,宁波315211 [3]国防科技大学计算机学院,长沙410073 [4]吉林大学商学与管理学院,长春130012

出　　处：《情报学报》2025年第3期353-368,共16页Journal of the China Society for Scientific and Technical Information

基　　金：国家社会科学基金青年项目“‘双碳’目标下‘技术-经济-区域’信息融合的创新生态系统构建及其协同演化研究”(22CTQ028)。

摘　　要：当前,全球科技创新呈现高速发展和高度融合的态势。准确识别出颠覆性技术主题以推动全面创新已成为科学技术发展和经济增长的关键动力。然而,传统的颠覆性技术主题识别方法主要依赖于单一模态数据,存在一定的局限性。本文基于CLIP(contrastive language-image pre-training)和LDAGV(linear discriminant analysis&global vectors for word representation)模型构建新闻文本与图像特征融合向量,通过k-means聚类迭代并结合3个颠覆性技术主题指标进行筛选,实现了多模态信息的融合以及主题的精准识别。以新能源领域为例,验证了该模型在颠覆性技术主题识别方面的可行性和有效性。与其他单一模态模型相比,多模态信息融合模型在颠覆性技术主题识别方面更具优势。Currently,global technological innovation is exhibiting a trend of rapid development and high integration.Accurately identifying disruptive technological themes that drive comprehensive innovation has become a key driving force for scientific and technological development and economic growth.However,traditional methods of subversive technology topic identification rely primarily on single-modal data,which have some limitations.This study constructs a news text and image feature fusion vector based on the CLIP(contrastive language-image pre-training)and LDAGV(linear discriminant analysis&global vectors for word representation)models,and uses k-means clustering iterations combined with three disruptive technology topic indicators for screening,achieving the fusion of multimodal information and accurate identification of topics.Using the new energy field as an example,the feasibility and effectiveness of the model for disruptive technology topic recognition are verified.Compared with other single-modal models,multimodal information fusion models have additional advantages in identifying disruptive technology topics.

关 键 词：颠覆性技术 主题识别 多模态融合 CLIP-LDAGV模型 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]