Text Style Transfer with Parameter-efficient LLM Finetuning and Round-trip Translation
This addresses the scarcity of parallel corpora for style mapping in TST, offering a practical solution for generating stylistically consistent text, though it is incremental as it builds on existing LLM and translation techniques.
The paper tackled the problem of Text Style Transfer (TST) by proposing a method that uses parameter-efficient fine-tuning of LLMs and round-trip translation to synthesize parallel datasets from monolingual corpora, achieving consistent superiority over zero-shot prompting and few-shot ICL techniques with improved BLEU and style accuracy scores across four domains.
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.