부산대학교 도서관

As the main information source in the process of clinical diagnosis and treatment, Traditional Chinese Medicine (TCM) electronic medical records contain information on symptoms, diagnosis, treatment methods, prescriptions and medicines, etc. Among them, the description of symptoms is flexible and variable, lacking uniform standards, and the extraction results of clinical symptom information are difficult t o b e recognized by the public, so an automatic extraction method of TCM clinical symptoms for electronic medical record is proposed. Firstly, the standard terminology of symptoms is used as the base corpus to recognize the symptoms in the standard terminology; then the deep learning method is used to extract the entities in the symptoms, which is divided into the recognition of backbone symptoms, acquisition methods and attributes; finally, the automatic extraction method is applied to the electronic medical record of a hospital for tympanites. The experimental results verified the scientific and operability of this method, and it was found that the electronic medical records of internal medicine and labeled 15% of TCM tympanites were used as the base corpus for symptom recognition using BERT-BiLSTM-CRF model (F1=95.94%), and symptom entity recognition using TextCNN model (F1=86.47%), while attribute recognition based on BERT-BiLSTM-CRF model (F1=93.52%) worked best. The extraction of TCM electronic medical records by this method can obtain standardized symptom information, which facilitates the deep utilization of symptom information and the mining of correlations between symptoms and diseases.