基于平行交互注意力网络的中文电子病历实体及关系联合抽取  

Parallel Interactive Attention Network Based Joint Entity and Relation Extraction for Chinese Electronic Medical Record

作　　者：李丽双[1] 王泽昊 秦雪洋 袁光辉 LI Lishuang;WANG Zehao;QIN Xueyang;YUAN Guanghui(School of Computer Science and Technology,Dalian University of Technology,Dalian,Liaoning 116024,China)

机构地区：[1]大连理工大学计算机科学与技术学院,辽宁大连116024

出　　处：《中文信息学报》2024年第6期108-118,共11页Journal of Chinese Information Processing

基　　金：国家自然科学基金(62076048);大连市科技创新基金(2020JJ26GX035)。

摘　　要：基于电子病历构建医学知识图谱对医疗技术的发展具有重要意义,实体和关系抽取是构建知识图谱的关键技术。该文针对目前实体关系联合抽取中存在的特征交互不充分的问题,提出了一种平行交互注意力网络(PIAN)以充分挖掘实体与关系的相关性,在多个标准的医学和通用数据集上取得最优结果;当前中文医学实体及关系标注数据集较少,该文基于中文电子病历构建了实体和关系抽取数据集(CEMRIE),与医学专家共同制定了语料标注规范,并基于该文所提出的模型实验得出基准结果。The construction of medical knowledge graph based on electronic medical records is of great significance to the development of medical technology,where entity and relation extraction plays a pivotal role.In this paper,we propose a Parallel Interactive Attention Network(PIAN)which can fully exploit the correlation between entity and relation.Since there are few Chinese medical entity and relation annotation datasets,we construct an entity and relation extraction dataset based on Chinese electronic medical records(CEMRIE),formulate the corpus annotation specification with medical experts,and give the benchmark results based on our proposed model.

关 键 词：实体关系联合抽取 双向特征交互模块 自注意力机制 中文电子病历 数据集标注与构建 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]