山西大学学报（自然科学版）2025，Vol.48Issue(6)：1059-1079,21.DOI:10.13451/j.sxu.ns.2025015
关系抽取方法研究综述
Review on Relation Extraction Methods Research
摘要
Abstract
With the ongoing evolution of information technology and deepening exploration in natural language processing(NLP),re-lation extraction,as a pivotal technique linking implicit connections among entities in vast textual data,plays a crucial role in intelli-gent question answering,knowledge graph construction,personalized information services,and etc.Amidst the surge of large lan-guage models,relation extraction technology has encountered unprecedented development opportunities,displaying substantial en-hancements in efficiency and intelligent level.In this paper,we first briefly introduce the background of the development of relation extraction,and then systematically review the development of relation extraction techniques,from the early methods relying on man-ual rules and knowledge bases,to the technological changes in traditional machine learning,and then to the development of deep learning-based relation extraction models.In addition,this paper mainly discusses the fusion application of large language models in the field of relational extraction.Finally,this paper not only discusses the challenges faced by current relation extraction research,such as limited data noise processing capability,insufficient capability for cross-domain generalization,and overlapping relation ex-traction,but also prospects the possible future research trends,including multimodal information fusion,document-level relation ex-traction,and innovations in unsupervised learning strategies,providing references to push forward the continuous advancement of the relation extraction technology.关键词
自然语言处理/关系抽取/大语言模型/多模态/文档级Key words
natural language processing/relation extraction/large language models/multi-modal/document-level分类
计算机与自动化引用本文复制引用
张勇,纪伟..关系抽取方法研究综述[J].山西大学学报（自然科学版）,2025,48(6):1059-1079,21.基金项目
深圳市科技计划项目(KJZD20230923114405012) （KJZD20230923114405012）
