«上一篇/Previous Article|本期目录/Table of Contents|下一篇/Next Article»

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:

34

期数:

2024年12期

页码:

1-8

栏目:

综述

出版日期:

2024-12-10

文章信息/Info

Title:

A Review of Event Relationship Extraction Methods

文章编号:

1673-629X(2024)12-0001-08

作者:

赵海宾1; 2; 3; 邵彩虹1; 2; 3; 李小龙1; 2; 3

1. 东华理工大学 测绘与空间信息工程学院,江西 南昌 330013;2. 东华理工大学 中核三维地理信息工程技术研究中心,江西 南昌 330013;3. 东华理工大学 自然资源部环鄱阳湖区域矿山环境监测与治理重点实验室,江西 南昌 330013

Author(s):

ZHAO Hai-bin1; 2; 3; SHAO Cai-hong1; 2; 3; LI Xiao-long1; 2; 3

1. School of Surveying and Geoinformation Engineering,East China University of Technology,Nanchang 330013,China;2. CNNC Engineering Research Center of 3D Geographic Information,East China University of Technology,Nanchang 330013,China;3. Key Laboratory of Mine Environmental Monitoring and Improving around Poyang Lake,Ministry of Natural Resources,East China University of Technology,Nanchang 330013,China

关键词:

自然语言处理; 事件关系抽取; 深度学习; 机器学习; 语义信息

Keywords:

natural language processing; event relationship extraction; deep learning; machine learning; semantic information

分类号:

TP391

DOI:

10.20165/j.cnki.ISSN1673-629X.2024.0271

摘要:

事件关系抽取以事件为基本语义单元从自然语言文本当中自动识别并提取出不同事件之间的逻辑关系。 事件关系抽取是自然语言处理与理解中的一个重要方向,在航空安全、金融、医学、舆情分析等领域有着很高的应用价值。 事件关系形式多样,目前事件关系抽取研究主要集中在时间关系、因果关系、共指关系和父子关系,该文以这四种关系作为分类标准介绍事件关系抽取任务。 首先,介绍并总结了四种关系抽取任务中常用的中英文数据集;然后,分别描述了四种关系抽取任务的概念和应用场景,详细介绍了各类关系抽取任务中的不同方法及其代表模型,在此基础上对比分析了各类方法的优缺点,并对各模型的实验性能和相关实验数据进行了归纳总结;最后,对当前事件关系抽取面临的研究难点进行了分析,针对未来的重点研究方向和发展趋势进行了展望,为进一步完善事件关系抽取方法提供了思路。

Abstract:

Event-relationship extraction,utilizing events as semantic units,automates the detection and extraction of logical connections between events in natural language texts. Event - relationship extraction is a critical direction in natural language processing and understanding,with significant applications in aviation security,finance,medicine,and public opinion analysis. Event relations come in many forms,and current research focuses primarily on temporal relations, causal relations, co - referential relations, and parent - child relations. We introduce the task of event relation extraction using these four relations as classification criteria. Firstly,we introduce and summarize the Chinese and English datasets commonly used in the four types of relationship extraction tasks. Then,we describe the concepts and application scenarios of the four types of relationship extraction tasks,introduce in detail the different methods and their rep-resentative models in each type of relationship extraction task,compare and analyze the strengths and weaknesses of each type of method on this basis, and summarize the experimental results and associated experimental data of each model; Finally, the current research problems of event - relationship extraction are reviewed, and the main research directions and development trends in the future are outlooked,providing recommendations for further enhancing the event-relationship extraction methods.

常用功能

更新日期/Last Update: 2024-12-10