NAST: A Non-Autoregressive Generator with Word Alignment for Unsupervised Text Style Transfer

TL;DR

NAST introduces a non-autoregressive approach with explicit word alignment for unsupervised text style transfer, improving content preservation, interpretability, and inference speed over traditional autoregressive models.

Contribution

The paper proposes a novel non-autoregressive generator that models word alignments to enhance style transfer quality and efficiency, addressing limitations of existing cycle-consistency based models.

Findings

Significantly improves style transfer performance.

Provides explainable word alignments.

Achieves over 10x inference speedup.

Abstract

Autoregressive models have been widely used in unsupervised text style transfer. Despite their success, these models still suffer from the content preservation problem that they usually ignore part of the source sentence and generate some irrelevant words with strong styles. In this paper, we propose a Non-Autoregressive generator for unsupervised text Style Transfer (NAST), which alleviates the problem from two aspects. First, we observe that most words in the transferred sentence can be aligned with related words in the source sentence, so we explicitly model word alignments to suppress irrelevant words. Second, existing models trained with the cycle loss align sentences in two stylistic text spaces, which lacks fine-grained control at the word level. The proposed non-autoregressive generator focuses on the connections between aligned words, which learns the word-level transfer between styles. For experiments, we integrate the proposed generator into two base models and evaluate them on two style transfer tasks. The results show that NAST can significantly improve the overall performance and provide explainable word alignments. Moreover, the non-autoregressive generator achieves over 10x speedups at inference. Our codes are available at https://github.com/thu-coai/NAST.