面向知识图谱构建的文档级关系抽取研究进展  

Knowledge graph-based document-level relation extraction:recent advances and perspectives

作　　者：曹倩倩 吴媛 Cao Qianqian;Wu Yuan(Information Technology Center,Taizhou University,Taizhou 318000,China;School of teacher Education(Physical Education),Taizhou University,Taizhou 318000,China)

机构地区：[1]台州学院信息技术中心,台州318000 [2]台州学院教师教育(体育)学院,台州318000

出　　处：《现代计算机》2024年第24期1-8,15,共9页Modern Computer

摘　　要：文档级关系抽取技术旨在通过融合句内和句间信息,捕获文档中多个关联实体指称所表达的复杂语义关系。准确提取出海量的非结构化数据中全部的“实体对”间关系是自下而上自动构建知识图谱的关键,也是下游任务顺利执行的基础。从基于图结构的方法与基于非图结构的方法两个方面概述了文档级关系抽取技术的研究进展,简要探讨其在知识图谱应用中的潜力,并展望未来研究的重点方向。研究发现,现有方法在文档级关系抽取方面取得了显著成果,对知识抽取乃至知识图谱自动构建至关重要。理清其发展现状并展望未来发展趋势将为后续相关研究提供清晰的思路。Document-level Relation extraction(DRE)technology aims to captures the complex semantic relations expressed by multiple entity references in the documents by integrating of both intra-sentence and inter-sentence information.Accurately extracting all the relations between entity pairs from massive unstructured data is not only the key to the automatic construction of knowledge graphs from the bottom to top,but also the basis for the smooth implementation of downstream tasks.In this paper,we summarize the research progress of DRE techniques using graph-based methods and non-graph-based methods.We also briefly introduce its potential applications in knowledge graphs,and explore the key directions of DRE’s future research.The existing methods have achieved remarkable results in DRE,highlighting its importance in knowledge extraction and automatic knowledge graph construction.We clarify the current development status and forecast future trends in DRE will provide direction for the follow-up studies.

关 键 词：文档级关系抽取 知识图谱 基于图结构的方法 基于非图结构的方法 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]