Adversarial feature matching for text generation

Published

Journal Article

© Copyright 2017 by the authors(s). 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 convolutional 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.

Duke Authors

Cited Authors

  • Zhang, Y; Gan, Z; Fan, K; Chen, Z; Henao, R; Shen, D; Carin, L

Published Date

  • January 1, 2017

Published In

  • 34th International Conference on Machine Learning, Icml 2017

Volume / Issue

  • 8 /

Start / End Page

  • 6093 - 6102

Citation Source

  • Scopus