SparseGAN: Sparse Generative Adversarial Network for Text Generation

TL;DR

SparseGAN introduces sparse, interpretable sentence representations inspired by sparse coding to improve the training stability and performance of GAN-based text generation models.

Contribution

It proposes a novel sparse coding approach for text representations in GANs, enabling fully differentiable training and enhanced sequence-level performance.

Findings

Improved BLEU scores across multiple datasets

More stable and efficient adversarial training process

Semantic-interpretable sentence representations

Abstract

It is still a challenging task to learn a neural text generation model under the framework of generative adversarial networks (GANs) since the entire training process is not differentiable. The existing training strategies either suffer from unreliable gradient estimations or imprecise sentence representations. Inspired by the principle of sparse coding, we propose a SparseGAN that generates semantic-interpretable, but sparse sentence representations as inputs to the discriminator. The key idea is that we treat an embedding matrix as an over-complete dictionary, and use a linear combination of very few selected word embeddings to approximate the output feature representation of the generator at each time step. With such semantic-rich representations, we not only reduce unnecessary noises for efficient adversarial training, but also make the entire training process fully differentiable. Experiments on multiple text generation datasets yield performance improvements, especially in sequence-level metrics, such as BLEU.