Journal of Big Data

Table 4 The performance of different \(\alpha\) parameter in the datasets

From: A semi-supervised short text sentiment classification method based on improved Bert model from unlabelled data

Model	Amazon Reviews		Chrome Reviews
Model	\(Accuracy \; (\%)\)	\(Macro \; F1 \; (\%)\)	\(Accuracy \; (\%)\)	\(Macro \; F1 \;(\%)\)
Text-CNN	85.332	85.168	88.900	91.476
LSTM	89.117	88.083	87.500	90.508
BiLSTM	90.626	90.130	90.040	91.716
Bert	91.025	89.480	91.900	92.520
Bert-MixMatchNL- Focal Loss	93.760	92.655	93.350	95.828

Bold values indicate the results of the model proposed in this paper

Back to article page