Hugging Face's logo

Datasets:
Tonykip
/
kalenjin-parallel-corpus

Tasks:
Translation
Modalities:
Text
Formats:
parquet
Languages:
Kalenjin
Swahili
English
Size:
10K - 100K
Tags:
kalenjin
low-resource-language
parallel-corpus
kenya
machine-translation
Libraries:
Datasets
pandas
Polars
License:
Dataset card Data Studio Files
xet
 Community
1

You need to agree to share your contact information to access this dataset

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

This corpus includes community-contributed Kalenjin translations and source text used under permission from the original data collection. By requesting access you agree to credit the sources listed in this card and to use the data responsibly. Contributor identifiers and recording metadata have been removed.

Log in or Sign Up to review the conditions and access this dataset content.

Kalenjin Parallel Corpus

A Kalenjin sentence-level parallel corpus — 89,020 pairs across English→Kalenjin and Swahili→Kalenjin, assembled from openly-licensed sources plus a permission-based community collection.

Config Direction Rows Source License
swa_kln Swahili → Kalenjin 18,326 thinkKenya/kenyan-low-resource-language-data (kln_swa) CC-BY-4.0
eng_kln English → Kalenjin 30,315 tketer/KalenjinEnglish (Bible, NIV) MIT
anvke_swa_kln Swahili → Kalenjin 25,861 Anv-ke community collection (used with permission) other / permission-based
anvke_eng_kln English → Kalenjin 18,503 Anv-ke community collection (used with permission) other / permission-based

Columns: source_text, target_text, source_lang, target_lang, domain, dialect, upstream_source, source_dataset, license.

Privacy

The Anv-ke records were originally a voice-collection dataset. Only the text pairs are included here — contributor identifiers (recorder_uuid), media paths (mediaPathId), and audio duration have been removed and appear nowhere in this corpus. Non-identifying metadata (domain, dialect, upstream_source) is retained.

Attribution

  • thinkKenya — Low-Resource Language Data (kln_swa), CC-BY-4.0 → swa_kln.
  • tketer/KalenjinEnglish — English NIV ↔ Kalenjin Bible, MIT → eng_kln.
  • Anv-ke community collection — used with permission → anvke_*. The Kalenjin side is community-translated; the source-language sentences come from these upstream corpora (see upstream_source): MedQuAD, EdTech Hub, Sprout, SALT, MAFAND-MT, OpenAfrica, KilimoSTAT, ANC/BBC Africa, Amazon MASSIVE, LRL Dataset, and others. Where an upstream source carries its own license or copyright, those terms also apply.

Access

This dataset is gated. It contains community-contributed data and source text whose upstream terms vary; access is granted per request so usage can be tracked and attributed. For redistribution, honor the per-source terms above.

Caveats

  • Registers differ — eng_kln is entirely Biblical (older ng' orthography in many rows); the others are mixed-domain. Weight accordingly.
  • The anvke_* configs preserve dialect labels (mostly NANDI) — useful, but not balanced across Kalenjin dialects.
  • Built by scripts/build_clean_open_corpus.py; see build_report.json for exact counts, duplicates removed, and the excluded-columns record.
Downloads last month
18
Total file size:
39.1 MB