Prompt-Based Editing for Text Style Transfer

TL;DR

This paper introduces a training-free, controllable prompt-based editing method for text style transfer that outperforms larger models in quality, using style classification and discrete word-level editing.

Contribution

The paper proposes a novel prompt-based editing approach transforming style transfer into a classification problem, enhancing control and efficiency without additional training.

Findings

Outperforms state-of-the-art systems with 20 times fewer parameters.

Effective in both automatic and human evaluations.

Demonstrates robustness across multiple benchmark datasets.

Abstract

Prompting approaches have been recently explored in text style transfer, where a textual prompt is used to query a pretrained language model to generate style-transferred texts word by word in an autoregressive manner. However, such a generation process is less controllable and early prediction errors may affect future word predictions. In this paper, we present a prompt-based editing approach for text style transfer. Specifically, we prompt a pretrained language model for style classification and use the classification probability to compute a style score. Then, we perform discrete search with word-level editing to maximize a comprehensive scoring function for the style-transfer task. In this way, we transform a prompt-based generation problem into a classification one, which is a training-free process and more controllable than the autoregressive generation of sentences. In our experiments, we performed both automatic and human evaluation on three style-transfer benchmark datasets, and show that our approach largely outperforms the state-of-the-art systems that have 20 times more parameters. Additional empirical analyses further demonstrate the effectiveness of our approach.