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

Table 3

The accuracy of the transfer learning classifier in the target domain question classification (%).

Model		A->B	A->C	A->D	B->E	B->A	AVG
AVG-T		51.16	52.39	49.17	77.60	55.89	57.24
Traditional transfer	TCA	55.02	51.42	49.02	75.36	57.14	57.59
	JDA	55.44	55.00	49.28	73.60	63.76	59.42
AVG-TT		55.23	53.21	49.15	74.48	60.45	58.50
Deep transfer	ULMFIT	63.68	63.66	52.50	83.58	66.14	65.91
	Flair	64.74	58.81	56.52	87.44	61.63	65.83
	XLNet	74.80	71.90	64.20	87.30	70.90	73.82
	BERT	77.30	77.20	68.40	92.40	74.90	78.04
	ALBERT	73.64	77.08	65.62	90.70	72.21	75.85
AVG-DT		70.83	69.73	61.45	88.28	69.16	71.89