摘要
基于症状体系识别的难点,提出一种创新的基于症状构成模式的非监督学习方法来实现电子病历症状实体的自动抽取,介绍其总体过程并与基于CRF序列标注的监督学习方法进行比较,试验证明本文所提出的方法具有良好的识别效果和可扩展性。
Based on difficulties in the recognition of the symptom system, the paper proposes an innovative unsupervised learning method based on the forming mode of symptoms so as to realize automatic extraction of symptomatic entities of Electronic Medical Records （EMR）. It introduces the overall process, and compares this method with the supervised learning method based on CRF sequence labe- ling. The test proves that the method proposed in this poper has better recognition effect and extendibility.
出处
《医学信息学杂志》
CAS
2016年第7期7-14,共8页
Journal of Medical Informatics
基金
上海市中医药事业发展三年行动计划(项目编号:ZY3-CCCX-2-1003)
国家高技术研究发展计划"心血管疾病与肿瘤疾病中西医临床大数据处理分析与应用研究"(项目编号:2015AA020107)
关键词
医疗实体抽取
症状构成模式
结构化电子病历
Medical entity extraction
Symptom composition pattern
Structured Electronic Medical Records
