深度学习基础上的中医实体抽取方法研究  

Study on the Entity Extraction Method of Traditional Chinese Medicine on the Basis of Deep Learning

作　　者：张艺品 关贝[1] 吕荫润 王翀[3] 吴炳潮 王永吉[1,2,3] 毕诗旋 ZHANG Yipin;GUAN Bei;LV Yinrun;WANG Chong;WU Bingchao;WANG Yongji;BI Shixuan(X-Lab Institute of Software, Chinese Academy of Sciences, Beijing 100190, China;University of Chinese Academy of Sciences, Beijing 100049;State Key Laboratory of Computer Science, Institute of Software, Chinese Academy of Sciences, Beijing 100190, China;Beijing University of Technology, Beijing 100124, China)

机构地区：[1]中国科学院软件研究所协同创新中心,北京100190 [2]中国科学院大学,北京100049 [3]中国科学院软件研究所计算机科学国家重点实验室,北京100190 [4]北京工业大学,北京100190

出　　处：《医学信息学杂志》2019年第2期58-63,共6页Journal of Medical Informatics

基　　金：科技部国家重点研发计划重点专项(项目编号:2017YFB1002300);大数据驱动的中医智能辅助诊断服务系统课题一"多模态异构中医药大数据高效获取与资源库建设"(项目编号:2017YFB1002301)和课题三"基于深度学习的中医多尺度认知方法和辩证论治分析模型"(项目编号:2017YFB1002303)

摘　　要：介绍命名实体识别及模型应用研究情况,以中医典籍作为数据源,采用深度学习方法,进行中医疾病、方剂、中草药等实体抽取,设计BiLSTM-CRF序列标注模型,构建中医典籍实验语料进行实验,结果表明该模型算法具有高度准确性。The paper introduces the study on named entity recognition and model application, conducts extraction of entities such as Traditional Chinese Medicine(TCM) diseases, prescription, Chinese herbal medicine, etc., by adoption of deep learning method and taking TCM classics as data sources, designs the model for sequencing tagging-BiLSTM-CRF. It also conducts experiments by building corpus of experiments in TCM classics. The result shows that the aforesaid model algorithm is of high accuracy.

关 键 词：知识图谱 实体抽取 中医 深度学习 

分 类 号：R-056[医药卫生]