Language Generation with Recurrent Generative Adversarial Networks without Pre-training
This addresses a bottleneck in text generation for NLP researchers, though it appears incremental as it builds on existing GAN methods with a specific training technique.
The paper tackled the problem of training GANs for language generation without pre-training by using recurrent neural networks and curriculum learning to generate sequences of increasing length, resulting in vastly improved quality compared to a convolutional baseline.
Generative Adversarial Networks (GANs) have shown great promise recently in image generation. Training GANs for language generation has proven to be more difficult, because of the non-differentiable nature of generating text with recurrent neural networks. Consequently, past work has either resorted to pre-training with maximum-likelihood or used convolutional networks for generation. In this work, we show that recurrent neural networks can be trained to generate text with GANs from scratch using curriculum learning, by slowly teaching the model to generate sequences of increasing and variable length. We empirically show that our approach vastly improves the quality of generated sequences compared to a convolutional baseline.