toiar/english-khasi-parallel-corpus-65k

English–Khasi Parallel Corpus (65k)

This dataset is a bilingual parallel corpus containing 65,328 sentence pairs for English and Khasi. It was constructed to support Machine Translation (MT) research and Natural Language Processing (NLP) development for Khasi, a low-resource Austroasiatic language spoken primarily in Meghalaya, India.

Dataset Summary

• Creator: Toiarbor Mawlieh
• Languages: English (en) ⇄ Khasi (kha)
• Total Pairs: 65,328

Dataset Structure

A typical row in the dataset contains a single parallel sentence pair:

{
  "english": "The fruit basket is full of apples and oranges.",
  "khasi": "Ka shang soh ka dap da ki soh-apul bad ki soh-ñiamtra."
}

Data Fields

• english: The English source text.
• khasi: The corresponding Khasi translation.

---

Limitations to Keep in Mind

• Missing Complex Words: This dataset is mostly general text. It does not have a lot of advanced medical, legal, or highly technical words.
• Short Sentences: The sentences are mostly short to medium length. Models trained on this might struggle to translate very long paragraphs accurately.
• Literal Translations: Some cultural idioms or metaphors may be translated literally rather than holding their deep cultural meaning.

---

Citation

If you use this dataset for a project or paper, please cite it like this:

@dataset{toiar_eng_kha_corpus_65K_2026,
  title     = {English-Khasi Parallel Dataset (65k)},
  author    = {Toiarbor Mawlieh},
  year      = {2026},
  publisher = {Hugging Face},
  license   = {CC BY-NC 4.0},
  url       = {https://huggingface.co/datasets/toiar/english-khasi-parallel-corpus-65k}
}