Skip to main content

Table 5 The effect of our baseline components on our development sets

From: A neural network approach to chemical and gene/protein entity recognition in patents

Model

CEMP Dev

GPRO Dev

Precision

Recall

F-score

â–³

Precision

Recall

F-score

â–³

Baseline

87.58

92.50

89.97

–

67.97

86.06

75.95

–

− Character embedding

86.27

90.98

88.56

− 1.41

66.67

83.69

74.22

− 1.73

− Capitalization feature

87.99

91.42

89.67

− 0.30

68.07

84.94

75.58

− 0.37

− CRF layer

84.84

88.55

86.66

− 3.31

62.81

79.41

70.14

− 5.81

− Post-processing

87.30

92.28

89.72

− 0.25

68.04

85.61

75.82

− 0.13

  1. Italic values denote the highest values