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

Table 4

Comparison of accuracy of each model in Chnsenticorp2000 data set and F1 score.

Model (threshold value of positive-negative probability ratio)	Accuracy	F1
BERT	0.9438	0.948
Dict-BERT(1.2)	0.9517	0.9520
Dict-BERT(1.4)	0.9517	0.9520
Dict-BERT(1.6)	0.9517	0.9520
Dict-BERT(1.8)	0.9508	0.9515
Dict-BERT(2.0)	0.9508	0.9515
Dict-BERT(3.0)	0.9500	0.9500