Text Style Transfer with Parameter-efficient LLM Finetuning and Round-trip Translation

TL;DR

This paper introduces a new approach for Text Style Transfer using parameter-efficient LLM fine-tuning combined with round-trip translation to generate parallel datasets from monolingual data, improving style transfer quality.

Contribution

It presents a novel method that synthesizes parallel data via roundtrip translation and employs efficient LLM fine-tuning, outperforming zero-shot and few-shot prompting techniques.

Findings

Outperforms zero-shot prompting in style transfer tasks.

Achieves higher BLEU and style accuracy scores.

Enhances robustness with retrieval-augmented generation.

Abstract

This paper proposes a novel method for Text Style Transfer (TST) based on parameter-efficient fine-tuning of Large Language Models (LLMs). Addressing the scarcity of parallel corpora that map between styles, the study employs roundtrip translation to synthesize such parallel datasets from monolingual corpora. This approach creates 'neutralized' text devoid of stylistic attributes, essentially creating a shared input style at training-time and inference-time. Experimental results demonstrate consistent superiority of this method over zero-shot prompting and fewshot ICL techniques measured by BLEU scores and style accuracy scores across four investigated domains. Furthermore, the integration of retrieval-augmented generation (RAG) for terminology and name knowledge enhances robustness and stylistic consistency.