基于实体类别信息的数据分析与关系抽取模型构建  

Data Analysis and Relation Extraction Model ConstructionBased on Entity Category Information

作　　者：杨航 张啸成 张永刚[1] YANG Hang;ZHANG Xiaocheng;ZHANG Yonggang(Key Laboratory of Symbol Computation and Knowledge Engineering of Ministry of Education,College of Computer Science and Technology,Jilin University,Changchun 130012,China)

机构地区：[1]吉林大学计算机科学与技术学院,符号计算与知识工程教育部重点实验室,长春130012

出　　处：《吉林大学学报(理学版)》2025年第2期428-436,共9页Journal of Jilin University:Science Edition

基　　金：吉林省自然科学基金(批准号:20200201447JC)。

摘　　要：针对文档级关系抽取任务中的实体多提及问题和实体对噪音问题,使用实体的类别信息,提出一个基于实体类别信息的关系抽取模型(EUT模型),该模型通过实体类别判断和类别对产生的关系类别先验两个子任务提高关系抽取结果.实体类别判断任务对实体进行类型标记后,再对实体所有提及进行类型分类,使实体的多个提及产生更丰富且相近的特征表示.关系类别先验任务使模型获得实体对的头尾类型所产生的关系分布先验,通过实体对的类别降低错误实体对噪音.为验证EUT模型的效果,在两个文档级数据集DocRED和Re-DocRED上进行实验,实验结果表明,该模型有效利用了实体的类型信息,与基础模型相比取得了更好的关系抽取效果,表明实体的类别信息对文档级关系抽取有重要影响.Aiming at the problem of multiple mentions of entities and the noise of entity pairs in the document-level relation extraction task,we proposed a relation extraction model(EUT model)based on entity type information.The model improved the relation extraction results through two sub-tasks:entity type judgment and a priori of the relation types produced by the type pairs.After the entity type judgment task labelled entities by type,then categorized all mentions of the entity by type,so that multiple mentions of the entity produced richer and similar feature representations.The relation category prior task enabled the model to obtain a prior of the relation distribution generated by the head and tail types of entity pairs,and reduced erroneous entity pair noise through the categories of entity pairs.In order to verify the effectiveness of the EUT model,the experiments were conducted on two document-level datasets,DocRED and Re-DocRED.The experimental results show that the model effectively utilizes the entity type information and achieves better relation extraction results compared to the base model,indicating that entity type information has an important impact on document-level relation extraction.

关 键 词：文档级关系抽取 知识图谱 结构化先验 自然语言处理 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]