文章摘要
李贞贞,胡思思,钟永恒,王辉,刘佳.基于AI Agent和文本多特征融合的前沿技术识别方法研究[J].数字图书馆论坛,2026,22(1):31~43
基于AI Agent和文本多特征融合的前沿技术识别方法研究
Cutting-Edge Technology Identification Method Based on AI Agent and Multi-Feature Text Fusion
投稿时间:2025-11-14  
DOI:10.3772/j.issn.1673-2286.2026.01.004
中文关键词: 前沿技术识别;AI Agent;多特征融合;Biterm主题模型
英文关键词: Cutting-Edge Technology Identification; AI Agent; Multi-Feature Fusion; Biterm Topic Model
基金项目:本研究得到湖北省技术创新专项软科学研究类重点项目“‘十五五’湖北基础研究重点方向任务及基础条件自主保障能力研究”(编号:2025EDA011)、青海省重点研发与转化计划“盐湖资源绿色高值利用科技成果数字化服务系统”(编号:2023-QY-213)资助。
作者单位
李贞贞 中国科学院武汉文献情报中心;科技大数据湖北省重点实验室 
胡思思 中国科学院武汉文献情报中心;科技大数据湖北省重点实验室 
钟永恒 中国科学院武汉文献情报中心;科技大数据湖北省重点实验室 
王辉 中国科学院武汉文献情报中心;科技大数据湖北省重点实验室 
刘佳 中国科学院武汉文献情报中心;科技大数据湖北省重点实验室 
摘要点击次数: 27
全文下载次数: 29
中文摘要:
      科技创新是产业高质量发展的强大驱动力,本研究提出一种基于人工智能体(AI Agent)和文本多特征融合的前沿技术识别方法,为把握技术突破方向、探寻技术创新机会提供借鉴。以基金项目、论文和专利数据为研究对象,首先,利用AI Agent进行数据预处理,提取文本关键词;其次,融合科技文献语义特征、分类特征和Biterm主题模型主题特征表征文本向量,增强数据特征,并利用聚类算法划分技术主题;最后,构建基于新颖性、增长性、影响力和创新性特征的前沿技术识别指标体系来识别前沿技术。对微发光二极管领域进行实证研究,融合语义、分类、主题3类特征的聚类方法表现出较好的效果,优于基于单一特征和两项特征的方法,并识别出芯片制造技术、全彩化技术等新兴前沿技术,发光材料与器件、显示屏技术、全息显示、医疗应用、检测与修复、驱动电路等热点前沿技术,以及巨量转移技术等潜在前沿技术,证明了方法的有效性和应用价值。
英文摘要:
      Technological innovation serves as a powerful driving force for the high-quality development of industries. This study proposes a method for identifying cutting-edge technologies based on AI agent and multi-feature text fusion, offering insights into grasping directions for technological breakthroughs and exploring opportunities for technological innovation. Using funding projects, academic papers, and patent data as research objects, this approach first applies AI agent for data preprocessing, extracting textual keywords. Then, the semantic features, classification features, and Biterm topic features of scientific and technological documents are fused to represent the text vector and enhance the data features. The clustering algorithm is used to divide the technical topics. Finally, a cutting-edge technology measurement index system based on novelty, growth, influence, and innovation is constructed to identify cuttingedge technologies. An empirical study is conducted in the field of micro-LED. Clustering methods that integrate semantic, classification, and topic features demonstrate superior performance compared to using only single or dual features. The analysis identifies emerging frontier technologies such as chip manufacturing technology and full-color display technology; highlights popular frontier technologies including light-emitting materials and devices, display technology, holographic display, medical applications, detection and repair, and driving circuits; and also uncovers potential frontier technologies like mass transfer technology. Thus, the effectiveness and application value of the proposed method are proved.
查看全文   查看/发表评论  下载PDF阅读器
关闭

分享按钮