Emotional Dialogue Generation Based on Conditional Variational Autoencoder and Dual Emotion Framework

Table 2

Objective evaluation with perplexity and accuracy.

Model	Perplexity	Accuracy
Seq2Seq [21]	67.2	0.205
ECM [14]	66.1	0.724
CVAE [16]	37.2	0.675
CVAE-MTDA [25]	34.6	0.692
EDGAN [27]	62.8	0.716
CVAE-DE	33.5	0.749