Skip to main content

Table 12 Class based accuracy of the top-five classes in the WE dataset

From: Toward multi-label sentiment analysis: a transfer learning based approach

Class

Model

XLNet

BERT

RF + LP

U.S.A.

98.31

98.42

92.67

Pinot Noir

97.74

98.03

95.75

Positive

93.53

93.15

88.60

Neutral

89.96

89.49

84.49

Negative

96.39

96.34

94.69