Table 7 Detailed cross-validation results for STRATEGY 2 and the HerBERT model.

From: Analysis of the retraining strategies for multi-label text message classification in call/contact center systems

k

Training data

Testing data

Accuracy

Emotica

Accuracy

Emotica

[%]

[%]

[%]

[%]

1

99.27

96.21

92.78

76.16

2

99.10

95.34

93.10

77.55

3

98.20

92.87

94.95

81.92

Average

98.86

94.81

93.61

78.55

SD

0.47

1.42

0.95

2.46

  1. Significant values are in bold.
Back to article page