Table 3.

The result of pre-trained models on Pitt dataset.

Model	Embedding	Classifier	Precision	Recall	Accuracy	F1
BertCNN	Bert	CNN	58.85	56.25	56.25	52.79
BertRCNN	Bert	RCNN	-	50.00	50.00	33.33
BertDPCNN	Bert	DPCNN	41.11	47.92	47.92	35.59
ERNIEDPCNN	ERNIE	DPCNN	-	50.00	50.00	33.33
BertLogistic	Bert	Logistic Regression	88	85	86.20	85.60
Transformer+GP	Transformer	Transformer	94.00	89.00	93.50	91.19