Deep Transfer Learning for Question Classification Based on Semantic Information Features of Category Labels

Table 5

The accuracy of L-ALBERT-FiT, ALBERT-fine-tune, and BERT-fine-tune in cross-domain question classification (%).

Learning rate	Train acc. (%)	Test acc. (%)
0.0001	93.73	67.18
0.0003	92.23	80.16
0.0005	87.11	84.7
0.0007	93.05	81.26
0.001	94.68	80.76
0.005	94.1	74.88
0.01	93.66	77.76
0.02	91.82	68.84