Table 3 The experimental results with different models on four datasets (accuracy in percentage).

LSTM⁷¹

49.27

77.68

65.70

90.54

fastText⁷⁹

57.70

75.14

61.38

92.81

CNN⁵¹

58.44

77.75

–

87.59

Text GCN¹

68.36

76.74

86.34

93.56

Text GNN¹

69.40

–

–

94.60

Tensor GCN⁶⁷

70.11

77.91

87.74

95.05

TextING^4,8,31

70.42

79.82

82.48

95.48

TextING-M^4,8,31

70.84

80.19

–

95.68

TG-Transformer³

70.40

–

–

95.20

HieGAT³⁷

69.84

78.04

85.84

94.54

RoBERTaGCN²¹

72.80

89.70

89.50

96.10

BertGAT²¹

71.20

86.50

87.40

96.50

RoBERTaGAT²¹

71.20

89.20

86.50

96.10

BertGCN²¹

72.80

86.00

89.30

96.60

ConTextING-BT³²

71.28

86.01

86.19

96.52

w.GAT-BT⁷⁸

71.51

86.16

86.25

96.28

ConTextING-RBT⁷⁸

72.53

89.43

85.00

96.40

w.GAT-RBT⁷⁸

72.06

89.24

84.97

96.15

MSABertGCN²²

74.72

86.97

-

96.96

Han-LT⁸⁰

71.46

–

87.62

–

HGCNN-MFF²⁸

71.22

79.82

88.32

-

CP-GNN⁷⁷

71.58

79.83

–

95.28

Our Model

73.34

83.50

91.50

96.10