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

Table 5

The results of model comparison on the two English datasets.

Models	SNIPS				StackOverFlow
		R	F1	Acc		R	F1	Acc
CNN	0.965	0.966	0.965	0.964	0.827	0.816	0.819	0.816
RNN	0.947	0.946	0.946	0.946	0.822	0.818	0.818	0.818
RCNN	0.966	0.967	0.966	0.966	0.851	0.839	0.842	0.839
RNN + Atten	0.961	0.961	0.961	0.961	0.845	0.827	0.831	0.827
Transformer	0.966	0.966	0.966	0.966	0.832	0.819	0.820	0.819
CNN + focal loss	0.969	0.969	0.969	0.969	0.838	0.831	0.832	0.831
RNN + focal loss	0.957	0.957	0.957	0.957	0.811	0.806	0.805	0.806
RCNN + focal loss	0.967	0.968	0.968	0.967	0.852	0.837	0.840	0.837
RNN + Atten + focal loss	0.970	0.970	0.970	0.97	0.819	0.805	0.808	0.805
Transformer + focal loss	0.971	0.970	0.970	0.97	0.824	0.815	0.816	0.815