首个针对乌克兰方言的LLM适配研究，解决低资源语言翻译难题。
Vuyko Mistral: Adapting LLMs for Low-Resource Dialectal Translation
- 构建9852对方言-标准语平行语料，用RAG生成52142条合成数据补足稀缺资源。
- 70亿参数微调模型在自动与LLM评估中均超越零样本GPT-4o。
- 适合低资源语言、方言保护及开放NLP研究者参考。
本文首次尝试将大语言模型（LLMs）适配至乌克兰语方言（以喀尔巴阡山区的胡丘尔方言为例），该方言资源匮乏且形态复杂。我们构建了包含9852组方言-标准乌克兰语句子对的平行语料库，并建立7320个方言词汇映射词典。为缓解数据不足问题，提出一种先进的检索增强生成（RAG）流水线，生成52142条合成平行翻译对，显著扩充数据集。我们使用LoRA微调多个开源LLM，在标准语到方言的翻译任务上进行评估，并与少样本GPT-4o进行对比。由于缺乏人工标注者，采用多指标评估策略，结合BLEU、chrF++、TER及基于GPT-4o的模型判断。结果表明，即使小型（7B）微调模型在自动指标和基于LLM的评估中也全面优于零样本基线如GPT-4o。所有数据、模型与代码已公开发布于：https://github.com/woters/vuyko-hutsul。
原文摘要 · Abstract (English)
In this paper we introduce the first effort to adapt large language models (LLMs) to the Ukrainian dialect (in our case Hutsul), a low-resource and morphologically complex dialect spoken in the Carpathian Highlands. We created a parallel corpus of 9852 dialect-to-standard Ukrainian sentence pairs and a dictionary of 7320 dialectal word mappings. We also addressed data shortage by proposing an advanced Retrieval-Augmented Generation (RAG) pipeline to generate synthetic parallel translation pairs, expanding the corpus with 52142 examples. We have fine-tuned multiple open-source LLMs using LoRA and evaluated them on a standard-to-dialect translation task, also comparing with few-shot GPT-4o translation. In the absence of human annotators, we adopt a multi-metric evaluation strategy combining BLEU, chrF++, TER, and LLM-based judgment (GPT-4o). The results show that even small(7B) finetuned models outperform zero-shot baselines such as GPT-4o across both automatic and LLM-evaluated metrics. All data, models, and code are publicly released at: https://github.com/woters/vuyko-hutsul
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。