Research on Named Entity Recognition of Chinese Electronic Medical Records Based on the BERT-BiLSTM-CRF Model

To address the problems of blurred entity boundaries, nested structures, and semantic ambiguity in Chinese electronic medical records (EMRs), this paper proposes a named entity recognition (NER) method based on the BERT-BiLSTM-CRF model. The model utilizes the BERT pre-trained language model to capture contextual semantic information, employs a Bidirectional Long Short-Term Memory (BiLSTM) network to further extract sequence features, and integrates a Conditional Random Field (CRF) layer for global label optimization, thereby improving the accuracy of medical entity recognition. The experimental data are derived from real clinical cases in the Anorectal Department of Xi'an Hospital of Traditional Chinese Medicine, comprising a total of 5,947 text records. After regular expression cleaning, word segmentation, and manual annotation, the dataset was divided into training, validation, and test sets in an 8:1:1 ratio. Experimental results show that the proposed model achieves superior performance in recognizing multiple categories of medical entities-such as diseases, symptoms, syndromes, and prescriptions-with an overall precision of 92.37%, recall of 94.25%, and F1-score of 93.29%. Compared with CNN-CRF and traditional BiLSTM-CRF models, the proposed approach demonstrates higher accuracy and robustness in Chinese EMR entity recognition.