Sentiment Classification Algorithm Based on the Cascade of BERT Model and Adaptive Sentiment Dictionary

Table 5

Comparison of accuracy of each model in Chnsenticorp4000 data set and F1.

Model (threshold value of positive and negative probability ratio)	Accuracy	F1
BERT	0.9508	0.9515
Dict-BERT(1.2)	0.9558	0.956
Dict-BERT(1.4)	0.9558	0.956
Dict-BERT(1.6)	0.9542	0.954
Dict-BERT(1.8)	0.9533	0.9535
Dict-BERT(2.0)	0.9508	0.9525
Dict-BERT(3.0)	0.9500	0.956