Multiple-Attribute Text Style Transfer

TL;DR

This paper challenges the necessity of disentangled latent representations for text style transfer and introduces a back-translation based model that effectively controls multiple attributes, improving generation quality on complex benchmarks.

Contribution

It proposes a novel multi-attribute text style transfer model that replaces disentanglement with back-translation, enabling better control and generation quality.

Findings

The model controls multiple attributes like gender and sentiment.

It outperforms disentanglement-based models on complex benchmarks.

It achieves a better balance between content preservation and style change.

Abstract

The dominant approach to unsupervised "style transfer" in text is based on the idea of learning a latent representation, which is independent of the attributes specifying its "style". In this paper, we show that this condition is not necessary and is not always met in practice, even with domain adversarial training that explicitly aims at learning such disentangled representations. We thus propose a new model that controls several factors of variation in textual data where this condition on disentanglement is replaced with a simpler mechanism based on back-translation. Our method allows control over multiple attributes, like gender, sentiment, product type, etc., and a more fine-grained control on the trade-off between content preservation and change of style with a pooling operator in the latent space. Our experiments demonstrate that the fully entangled model produces better generations, even when tested on new and more challenging benchmarks comprising reviews with multiple sentences and multiple attributes.