电子病历中命名实体的智能识别

电子病历中命名实体的识别对于构建和挖掘大型临床数据库以服务于临床决策具有重要意义,而我国目前对此的研究相对较少。在比较现有的实体识别方法和模型后,采用条件随机场模型(CRF)机器学习的方法,对疾病、临床症状、手术操作3类中文病历中常见的命名实体进行智能识别。首先,通过分析电子病历的数据特征,选择以语言符号、词性、构词特征、词边界、上下文为特征集。然后,基于随机抽取的来自临床医院多个科室的电子病历数据,构建小规模语料库并进行标注。最后,利用条件随机场算法执行工具CRF++进行3次对照实验。通过逐步分析特征集中的多种特征对CRF自动识别的影响,提出在中文病历环境下CRF特征选择和模板设计的一些基本规则。在对照实验中,本方法取得了良好效果,3类实体的最佳F值分别达到了92.67%、93.76%和95.06%。

Intelligent Recognition of Named Entity in Electronic Medical Records

The named entity recognition in electronic medical records is very important for building and mining large-scale clinical data to serve the clinical decision-making.However,in China,there are few relative studies on this.In comparison to the existing entity recognition methods and models,this paper attempted to use a machine learning method based on conditional random field(CRF) model to intelligently recognize three common types of the named entity in Chinese medical records,they are diseases,clinical symp...