Joint Character-Level Convolutional and Generative Adversarial Networks for Text Classification

Table 3

Dataset statistics.

Dataset	Before extracting
	Train	Test	Unlabeled
AG-News	120 k	7.6 k	—	46	4	5
DBPedia	560 k	70 k	—	56	14	5
20NG	11314	7532	—	221	20	10
IMDB	25 k	25 k	50 k	239	2	10

is the number of categories, is the average length of a document, and is the extracting times of the sample sets.