Table 1 Performance comparison of AFIENet combined with different backbone network on THUCNews, ChnSentiCorp, and ChemicalPro datasets (%).

From: Adaptive feature interaction enhancement network for text classification

Method

THUCNews

ChnSentiCorp

ChemicalPro

Average

Accuracy

F1

Accuracy

F1

Accuracy

F1

Accuracy

F1

TextCNN

91.02

91.03

85.51

85.30

86.45

86.44

87.66

87.59

AFIE-TextCNN

97.55

97.55

88.64

88.65

88.24

88.20

91.48

91.47

DPCNN

91.15

91.12

84.93

84.83

85.72

85.63

87.27

87.19

AFIE-DPCNN

93.62

93.59

86.05

86.09

87.86

87.82

89.18

89.17

LSTM

90.99

91.00

83.52

83.38

81.88

82.01

85.46

85.46

AFIE-LSTM

91.60

91.56

85.45

85.12

84.52

84.42

87.19

87.03

Transformer

89.73

89.74

82.42

81.90

78.59

78.49

83.58

83.38

AFIE-Transformer

90.01

90.01

83.13

82.87

81.37

81.29

84.84

84.72

  1. AFIE-* indicates that both GE-Net and LA-Net in AFIE-Net use * as the feature extractor.
  2. Models * and AFIE - * form a comparison group. The bold is used to mark the group with better experimental performance in this comparison.