Disentangled Representation Learning for Non-Parallel Text Style Transfer

TL;DR

This paper introduces a method for disentangling style and content in language models using auxiliary tasks, leading to improved non-parallel text style transfer with better accuracy, content preservation, and fluency.

Contribution

It presents a novel disentangled representation learning approach with auxiliary multi-task and adversarial objectives for non-parallel text style transfer.

Findings

Achieves higher transfer accuracy than previous methods.

Improves content preservation in style transfer.

Enhances language fluency in generated text.

Abstract

This paper tackles the problem of disentangling the latent variables of style and content in language models. We propose a simple yet effective approach, which incorporates auxiliary multi-task and adversarial objectives, for label prediction and bag-of-words prediction, respectively. We show, both qualitatively and quantitatively, that the style and content are indeed disentangled in the latent space. This disentangled latent representation learning method is applied to style transfer on non-parallel corpora. We achieve substantially better results in terms of transfer accuracy, content preservation and language fluency, in comparison to previous state-of-the-art approaches.