Question Answering System of Traditional Chinese Medicine based on BERT and Knowledge Graph

With the continuous growth in demand for health information and medical services, there is a large number of medical question-answering systems available in the market. The system has made progress in the realm of general medicine, but still faces challenges in the field TCM (Traditional Chinese Medicine). For example, problems such as inaccurate information and lack of personalization of answers may arise. There is a vast amount of TCM knowledge available online for medical diagnosis, but traditional search engines are unable to make judgments based on the actual condition of patients, failing to meet user needs. Therefore, this paper proposes a TCM question and answer system based on BERT and knowledge graph. The experimental results show that both the BERT+BiLSTM+ CRF model and the BiLSTM+Attention model have achieved improvements, enhancing the precision, recall, and F1-score of TCM information recognition. This ensures the question-answering system's responses are more closely aligned with user needs.