An Integrated Deep Generative Model for Text Classification and Generation

Table 2

The classification error rate (%) on the IMDB dataset of each method. Lower is better.

Methods	Error rate on IMDB
SVM + bow(baseline)	11.55
WRRBM+bow[30]	10.77
SVM+bow2+NB[31]	8.78
NB-LM + bow3[32]	8.13
LSTM	13.50
VAE+NN(ours)	8.86