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

Table 4

Test performance on the AG-news classification task.

Method	Dataset size
	Part-1	Part-2	Part-3	Part-4	Part-5	Part-6	Part-7	Part-8	Total
Tree-LSTM	37.96	58.13	68.33	76.20	87.12	90.75	91.56	91.80	91.83
Self-Attentive	30.81	54.89	72.26	85.47	86.37	90.83	91.28	92.03	91.17
Emb-CNN	48.83	68.38	79.74	84.11	83.85	85.94	87.93	89.36	90.08
Char-CNN	52.97	72.14	78.93	84.46	85.85	88.02	87.36	87.75	87.28
Char-CRNN	52.47	68.36	75.29	85.17	84.78	90.26	91.53	91.26	91.44
FastText	53.36	68.48	75.48	80.83	85.37	91.05	91.25	91.54	91.51
L-MIXED	24.68	25.45	66.21	79.25	84.20	90.27	93.82	93.90	94.38
DPCNN	25.94	25.62	69.85	77.29	82.44	86.71	90.89	91.12	93.13
LEAM	33.24	49.13	61.85	70.21	80.97	87.38	91.75	91.82	92.45
Ad-Training	47.29	57.63	69.15	76.21	82.37	87.83	90.89	90.22	91.37
Text GCN	57.32	68.96	77.83	84.54	87.62	89.35	90.74	91.32	92.56
CCNN-GAN	66.47	77.53	85.37	86.93	88.34	91.48	91.25	91.79	91.94