Khakas (kjh) version of SmolSent and SmolDoc
Hello!
My name is Vasily Adeshkin, and I am a researcher at the Khakas Research Institute for Language, Literature, and History.
I have been following the release of the Google SMOL dataset. To support machine translation research for under-resourced Siberian languages, we have translated the SMOL dataset into the Khakas language (kjh), a member of the Siberian subgroup of the Turkic language family.
In this Pull Request, we submit the translated files for both the sentence-level (smolsent) and document-level (smoldoc) subsets in Khakas, aligned with English and Russian and matching the original SMOL structure.
This dataset is also published on Hugging Face here:
👉 https://huggingface.co/datasets/adeshkin/google-smol-en-ru-kjh
If you are planning to expand the official google/smol dataset to include more low-resource languages, we would like to propose integrating this Khakas translation.
Additionally, we have compiled other related resources for Khakas, including a Khakas-Russian parallel corpus (159k sentences), a monolingual corpus (258k sentences), and a translation of the FLORES+ dataset. They are all collected here:
👉 adeshkin/khakas-translation
Please let us know if you would be interested in integrating this data, or if there are any guidelines we should follow to assist with a contribution.
Thank you for your dedication to low-resource language translation.
Best regards,
Vasily Adeshkin