Skip to main content

Table 3 Models accuracy results on the Arabic hotel reviews dataset, Arabic news dataset, and HAAD

From: Arabic aspect sentiment polarity classification using BERT

Models

Datasets

Arabic hotel reviews

Arabic news

HAAD

Baseline

76.4

61.47

29.70

LSTM

81.49

74.81

52.30

TD-LSTM

81.79

76.62

54.26

AB-LSTM-PC

82.60

77.11

58.08

INSIGHT-1 (CNN)

82.71

–

–

HP-LSTM

82.80

–

–

IAN-BGRU

83.98

78.78

68.62

MBRA

87.31

–

–

BERT-linear-single

85.93

82.12

68.25

BERT-linear-pair

89.51

85.73

73.23