Research Article

An Integrated Deep Generative Model for Text Classification and Generation

Table 3

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

Methods Error rate on SST1 Error rate on SST2

SVM + bow(baseline) 61.4 22.8
RAE[33] 56.8 17.6
MV-RNN[34] 55.6 17.1
RNTN[35] 54.3 14.6
DCNN[13] 51.5 12.2
VAE+NN(ours) 55.3 16.9