A Hybrid Neural Network BERT-Cap Based on Pre-Trained Language Model and Capsule Network for User Intent Classification

Table 3

The performance of different methods on two English datasets.

Models\datasets	SNIPS				StackOverFlow
		R	F1	Acc		R	F1	Acc
BERT	0.970	0.972	0.970	0.970	0.878	0.872	0.872	0.872
BERT + CNN	0.970	0.971	0.970	0.970	0.874	0.873	0.873	0.873
BERT-Cap	0.977	0.978	0.977	0.977	0.882	0.880	0.880	0.880
BERT-Cap + focal loss	0.978	0.980	0.979	0.979	0.883	0.878	0.879	0.878