Skip to main content

Table 10 Results of different NER models on cEHRNER

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

Models

Precision

Recall

F1

Cost time

BiLSTM-CRF

84.72

85.14

84.92

5 h 18  m

RD + CNN + CRF [31]

86.36

87.23

86.79

5 h

ELMO-BiLSTM-CRF [32]

87.54

87.33

87.43

4 h 37 m

BERT-BiLSTM-CRF

89.35

89.96

89.65

3 h 31 m

BERT-wwm + BiLSTM + CRF [33]

90.17

90.46

90.31

3 h 54 m

MC_BERT-BiLSTM-CRF

91.69

92.14

91.91

3 h 25 m

MC_BERT-BiLSTM-MHA-CRF

92.11

92.32

92.21

3 h 36 m

MC_BERT-BiLSTM-CNN-CRF

92.24

92.54

92.38

3 h 47 m

MC_BERT-BiLSTM-CNN-MHA-CRF

92.78

92.88

92.82

3 h 56 m