A BERT-Based Named Entity Recognition Method of Warm Disease in Traditional Chinese Medicine

Traditional Chinese medicine (TCM) documents have been handed down through the ages, containing rich theoretical knowledge and clinical experience. These unstructured data are the foundation for building the digital knowledge system of TCM. However, written in ancient Chinese, the TCM books have complex grammatical rules and terms which are different from modern medicine, inducing difficulty in entity annotation and recognition. In order to solve the problem of lacking labeled data, we construct a dataset with Wenbing Tiaobian, a classic work of TCM on the warm disease, identify six entities and annotate the book with the BIOES method. The BERT-BILSTM-CRF model is used to conduct experiments on the dataset with an F1 value of 91.4%. The results verify the effectiveness of the model in NER tasks and advance the construction of knowledge graphs in TCM.