An Extended Text Combination Classification Model for Short Video Based on Albert

Table 3

Comparison on the test set between the multitext fusion model based on attention mechanism and other Bert-based composite models trained by supervised learning and semisupervised learning methods.

Model	Val Acc	Val loss	Test Acc	Test loss	Supervised/semisupervised learning
Ours	94.08%	0.2	93.95%	0.19	Semisupervised
Albert	93.66%	0.29	91.63%	0.21	Semisupervised
Albert	92.97%	0.29	82.88%	0.44	Supervision
Albert _CNN	90.93%	0.3	80.31%	0.51	Supervision
Albert _DPCNN	91.93%	0.26	82.54%	0.47	Supervision
Albert _RNN	91.41%	0.3	82.22%	0.47	Supervision
ERNIE	90.16%	0.26	80.66%	0.51	Supervision