Adversarial feature matching for text generation

TLDR

GANs have succeeded in generating realistic synthetic data, but convergence problems and challenges with discrete data limit their use for text generation. The authors propose a framework to generate realistic text through adversarial training. They use an LSTM generator and a convolutional discriminator, and instead of the standard GAN objective, they match high‑dimensional latent feature distributions of real and synthetic sentences using a kernelized discrepancy metric. This approach eases adversarial training by reducing mode collapse, and experiments demonstrate superior quantitative performance and realistic‑looking sentence generation.

Abstract

The Generative Adversarial Network (GAN) has achieved great success in generating realistic (real-valued) synthetic data. However, convergence issues and difficulties dealing with discrete data hinder the applicability of GAN to text. We propose a framework for generating realistic text via adversarial training. We employ a long short-term memory network as generator, and a con-volutional network as discriminator. Instead of using the standard objective of GAN, we propose matching the high-dimensional latent feature distributions of real and synthetic sentences, via a kernelized discrepancy metric. This eases adversarial training by alleviating the mode-collapsing problem. Our experiments show superior performance in quantitative evaluation, and demonstrate that our model can generate realistic-looking sentences.