Research Article

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\datasetsSNIPSStackOverFlow
RF1AccRF1Acc

BERT0.9700.9720.9700.9700.8780.8720.8720.872
BERT + CNN0.9700.9710.9700.9700.8740.8730.8730.873
BERT-Cap0.9770.9780.9770.9770.8820.8800.8800.880
BERT-Cap +focal loss0.9780.9800.9790.9790.8830.8780.8790.878