Skip to main content

Table 6 Performance comparison of BiLSTM-Att-CRF model and basic models

From: An attention-based deep learning model for clinical named entity recognition of Chinese electronic medical records

 

Dataset of CCKS 2018

Dataset of CCKS 2017

 

Precision

Recall

F-score

Precision

Recall

F-score

CRF

85.63%

79.58%

82.49%

87.32%

83.06%

85.14%

+POS

85.94%

81.04%

83.42%

88.97%

85.12%

87.01%

+Dic

88.98%

82.82%

85.79%

90.34%

86.06%

88.15%

+POS + Dic

88.72%

83.57%

86.04%

91.26%

87.11%

89.14%

BiLSTM-CRF

85.20%

83.09%

84.13%

90.09%

89.24%

89.66%

+POS

85.23%

82.73%

83.96%

89.36%

89.74%

89.55%

+Dic

85.62%

83.34%

84.46%

90.17%

90.44%

90.30%

+POS + Dic

86.02%

82.93%

84.45%

90.31%

90.12%

90.22%

BiLSTM-Att-CRF

86.51%

84.38%

85.43%

90.11%

90.47%

90.29%

+POS

86.62%

84.36%

85.48%

90.33%

89.88%

90.10%

+Dic

86.97%

84.79%

85.87%

89.87%

90.75%

90.31%

+POS + Dic

87.09%

85.13%

86.11%

90.41%

90.49%

90.48%

  1. The bold values denote the highest values