Saved in:
Bibliographic Details
Main Authors: Kyslyi, Roman, Maksymiuk, Yuliia, Pysmennyi, Ihor
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2506.07617
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911079053918208
author Kyslyi, Roman
Maksymiuk, Yuliia
Pysmennyi, Ihor
author_facet Kyslyi, Roman
Maksymiuk, Yuliia
Pysmennyi, Ihor
contents 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
format Preprint
id arxiv_https___arxiv_org_abs_2506_07617
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Vuyko Mistral: Adapting LLMs for Low-Resource Dialectal Translation
Kyslyi, Roman
Maksymiuk, Yuliia
Pysmennyi, Ihor
Computation and Language
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
title Vuyko Mistral: Adapting LLMs for Low-Resource Dialectal Translation
topic Computation and Language
url https://arxiv.org/abs/2506.07617