基于BERT的电子病历实体关系联合抽取研究  

Study on Joint Extraction of Chinese Electronic Medical Record Entity Relationship Based on BERT

作　　者：黄晓芳[1] 陈剑秋 周祖宏[2] 廖敏[2] HUANG Xiaofang;CHEN Jianqiu;ZHOU Zuhong;LIAO Min(School of Computer Science and Technology,Southwest University of Science and Technology,Mianyang 621000,China;Mianyang Central Hospital,Mianyang 621000,China)

机构地区：[1]西南科技大学计算机科学与技术学院,绵阳621000 [2]绵阳市中心医院,绵阳621000

出　　处：《医学信息学杂志》2023年第2期28-34,共7页Journal of Medical Informatics

基　　金：四川省科技厅重点研发项目“面向多语种的全球前沿技术智能汇集与应用平台”(项目编号:2021YFG0031)。

摘　　要：分析中文电子病历数据实体关系提取常用方法,提出一种基于双向编码器表征的实体关系联合抽取算法,使用级联解码器以及指针标注方法完成实体关系抽取及实体识别,实验结果证明该方法可有效抽取电子病历实体关系。The paper analyzes the common methods of entity relationship extraction for Chinese electronic medical record(EMR) data, proposes a joint extraction algorithm of entity relationship based on bidirectional encoder representation from transformers(BERT), which uses cascade decoder and pointer annotation method to complete entity relationship extraction and entity recognition. Experimental results show that the algorithm used in the paper can effectively extract the entity relationship of electronic medical record.

关 键 词：电子病历 关系抽取 联合抽取模型 自然语言处理 

分 类 号：R-058[医药卫生]