New RT-SFT method enables text style transfer without parallel corpora

Researchers have developed a novel method called RT-SFT (Roundtrip Translation Style Transfer) to perform text style transfer without requiring parallel corpora. This technique leverages the inherent style-stripping capabilities of neural machine translation systems. By roundtrip translating a monolingual corpus through a pivot language, a pseudo-parallel dataset is generated, which is then used to fine-tune an instruction-tuned LLM as a stylizer. This approach has demonstrated superior performance compared to state-of-the-art methods across four different style domains. AI

IMPACT This method could significantly reduce the data requirements for developing text style transfer models, making the technology more accessible.

RANK_REASON The cluster contains an academic paper detailing a new method for text style transfer. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• alphaXiv
• arXiv
• DagsHub
• Hugging Face
• LoRA+
• RT-SFT
• Ruoxi Liu
• ScienceCast

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →