Aiming at the problems of inaccurate boundary division and low entity recognition rate in the Named Entity Recognition (NER) task of Chinese Electronic Medical Records (EMR), the paper proposes a CNN-BiLSTM-CRF model based on deep learning, expounds the structure and principle of the model in detail, and collects 3 127 Chinese EMR for experiments to verify the performance of the model. The results show that this model achieves better recognition effect and better performance.