A Traditional Chinese Medicine Terminology Recognition Model Based on Deep Learning: A TCM Terminology Recognition Model

Although the classics of traditional Chinese medicine are extensive and profound, the application and research of the deep learning model in traditional Chinese medicine terminology recognition stay scarce. In order to make use of the knowledge in these classics, a traditional Chinese medicine terminology recognition model named BERT-BiLSTM-CRF is presented and achieve superior performance as shown in the Results section. The design process of the BERT-BiLSTM-CRF model combines the transfer learning strategy, the pre-training language model, and the classical machine learning model. Specifically, the semantic features of TCM sample sequences is first extracted by transferring a BERT model that pre-trained on other large-scale Chinese corpora. Then go to the BiLSTM module to abstract the semantic features of sequence context. And last, introduce CRF to learn the transfer features between context tags. In the experiment, the BERT-BiLSTM-CRF model is compared with a variety of benchmark models and outperforms others.