Named entity recognition of Chinese electronic medical records based on a hybrid neural network and medical MC-BERT

Table 8 Results of different NER models on CCKS-2017

Models	Precision	Recall	F1	Cost time
BiLSTM-CRF	89.47	89.06	89.26	6 h 41 m
RD-CNN-CRF [31]	90.63	92.02	91.32	5 h 56 m
ELMO-BiLSTM-CRF [32]	91.48	93.92	92.66	5 h 41 m
BERT-BiLSTM-CRF	92.03	94.46	93.23	4 h 35 m
BERT-WWM + BiLSTM + CRF [33]	92.24	94.74	93.47	4 h 46 m
MC_BERT-BiLSTM-CRF	92.25	94.98	93.60	4 h 22 m
MC_BERT-BiLSTM-MHA-CRF	92.62	94.89	93.74	4 h 33 m
MC_BERT-BiLSTM-CNN-CRF	92.73	95.11	93.90	4 h 41 m
MC_BERT-BiLSTM-CNN-MHA-CRF	93.04	95.43	94.22	4 h 57 m

ISSN: 1472-6947