Jendersen/welsh-breton-cornish-filtered-n-readied

Welsh-Breton-Cornish Filtered & Readied

Dataset Description

This dataset provides parallel text data for machine translation from English into three Celtic languages:

• —Breton (br)
• —Welsh (cy)
• —Cornish (kw)

It consists of prompt–translation pairs: the text field contains an English instruction with a target language specifier, and the target field contains the corresponding translation in that language.

The data has been filtered and prepared to support training and evaluation of translation models.

Features

The {target_language} part of the prompt indicates which language the target translation is in.

Languages

• —Breton (br)
• —Welsh (cy)
• —Cornish (kw)

Dataset Splits

Dataset Size

License

This dataset is released under the MIT License. ---

Task Categories

• —Translation (English → Breton / Welsh / Cornish) ---

Example

{
  "target": "Mes mars yw avlan tir agas perghennogeth, treusewgh dhe dir perghennogeth an Arloedh le may powes tabernakel an Arloedh, ha kemmerewgh rann ragowgh agas honan yn agan mysk. Saw na rebellyewgh erbynn an Arloedh na er agan pynn ni, ow trehevel ragowgh alter nag yw alter an Arloedh agan Duw.",
  "text": "translate English to Cornish: But if you consider the land of your possession to be unclean, cross over to the land in which is the tabernacle of the Lord, and live among us. But do not withdraw from the Lord, and from our fellowship, by building an altar contrary to the altar of the Lord our God."
}

Usage

Load the dataset with the Hugging Face datasets library:

from datasets import load_dataset

dataset = load_dataset("Jendersen/welsh-breton-cornish-filtered-n-readied")

# Access training split
train_data = dataset["train"]

# Access test split
test_data = dataset["test"]

Citation

If you use this dataset in your work, please cite it as:

@dataset{jendersen_welsh_breton_cornish,
  author       = {Jendersen},
  title        = {Welsh-Breton-Cornish Filtered \& Readied},
  year         = {2026},
  publisher    = {Hugging Face},
  url          = {https://huggingface.co/datasets/Jendersen/welsh-breton-cornish-filtered-n-readied}
}

Notes

• —The dataset is ready-to-use for translation tasks.
• —It has been filtered to remove low‑quality or noisy examples.
• —The prompt format allows the model to infer the target language from the instruction, making it suitable for multilingual translation setups.