基于细粒度语义的领域知识流构建方法研究 ——以人工智能领域为例  

Research on the Construction Method of Domain Knowledge Flow Based on Fine-Grained Semantics:A Case Study of the Field of Artificial Intelligence

作　　者：叶梓萌 钱力[1,2,3] 丁洁兰 刘志博[1,2] Ye Zimeng;Qian Li;Ding Jielan;Liu Zhibo(National Science Library,Chinese Academy of Sciences,Beijing 100190;Department of Information ResourcesManagement,School of Economics and Management,University of Chinese Academy of Sciences,Beijing 100190;Key Laboratory of New Publishing and Knowledge Services for Scholarly Journals,Beijing 100190)

机构地区：[1]中国科学院文献情报中心,北京100190 [2]中国科学院大学经济与管理学院信息资源管理系,北京100190 [3]国家新闻出版署学术期刊新型出版与知识服务重点实验室,北京100190

出　　处：《情报理论与实践》2025年第4期143-151,共9页Information Studies:Theory & Application

基　　金：国家社会科学基金重大项目“大数据驱动的科技文献语义评价体系研究”(项目编号:21&ZD329);中国科学院文献情报能力建设专项项目“科技态势感知与分析能力建设”(项目编号:E3290909)的成果。

摘　　要：[目的/意义]基于细粒度语义的视角,通过研究问题、方法抽取与领域知识流状态识别以刻画科研领域的知识发展脉络,把握领域知识的发展趋势。[方法/过程]验证GPT提示学习的实体抽取方法的有效性,探索GPT提示学习在实体抽取中的应用效果,并创新性地提出基于特征向量中心性与Z-score的重点实体识别方法。同时,提出一套领域知识流的量化方式、知识流状态的种类及识别方法,以对领域知识流识别与进一步分析。[结果/结论]将领域知识流动状态划分为知识新生、知识衰亡、知识继承、知识合并与知识分裂5种状态,实现对领域知识主题的有效识别与刻画。研究成果表明,此方法能够有效揭示人工智能研究领域中的知识脉络,是领域画像的有力工具。[Purpose/significance]From the perspective of fine-grained semantics,this study explores the extraction of research problems and methods and the identification of domain knowledge flow states to trace the knowledge evolution in research fields and grasp the development trends of domain knowledge.[Method/process]The study validates the effectiveness of entity extraction using GPT prompt learning and examines its application in entity extraction.An innovative method is proposed for identifying key entities based on eigenvector centrality and Z-score.Additionally,a quantitative framework for domain knowledge flows is introduced,along with the categorization and identification methods for knowledge flow states,enabling in-depth analysis of domain knowledge flows.[Result/conclusion]The domain knowledge flow states are categorized into five types:knowledge emergence,knowledge obsolescence,knowledge inheritance,knowledge merging,and knowledge splitting.This approach achieves effective identification and characterization of domain knowledge themes.The findings demonstrate that the proposed method effectively reveals the knowledge evolution in the field of artificial intelligence,making it a powerful tool for domain profiling.

关 键 词：领域知识流 细粒度语义 提示学习 领域画像 语义挖掘 知识流动状态 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]