Open guotong1988 opened 5 years ago
It is addressed in the paper in the introduction . Language models learnt in traditional maximum likelihood approach suffer exposure bias. Though there are techniques like scheduled sampling to overcome it, they do not address the problem at its roots. here an adversarial training beats that problems
I'm new to LeakGAN or SeqGAN or TextGAN. I know GAN is to generate text and let discriminator un-judge-able to real text and gen-text. LM(language model) is the task of predicting the next word and can also be used to generate text.