基于深度学习的电子病历命名实体识别及其在知识发现中的应用

Named Entity Recognition from Electronic Medical Records Based on Deep Learning and its Application in Knowledge Discovery

通过引入医学文本语言和文档类别特征,构建了一个基于深度学习的电子病历命名实体识别系统。识别的实体包括身体部位、症状和体征、疾病和诊断、检查和检验以及治疗5大类。基于模型识别的结果,将其应用在基于共现的临床知识发现中。命名实体识别系统的准确率为93.29%,召回率为93.53%,F1值为93.41%。医学语言特征的引入能够进一步提高基于深度学习的医学实体识别系统的效果,实体识别的结果可以作为电子病历知识发现的基础。

By introducing medical text language and document category features, a named entity recognition system for electronic medical records based on deep learning technology was constructed. Identified entities include body parts, symptom, disease and diagnosis, examination and treatment. The entity recognition results were applied in a knowledge discovery system based on co-occurrence. The accuracy of the system is 93.29%, the recall rate is 93.53%, and the F1 score is 93.41%. The introduction of medical language features can further improve the effect of medical entity recognition system based on deep learning, and the extracted results can be utilized as the basis of knowledge discovery from electronic medical records.