ghananlpcommunity/ghana-corpus

This dataset is shared under CC BY-NC 4.0, which means you are free to use, share, and adapt it for non-commercial research and educational purposes with attribution. You can read the full license at https://creativecommons.org/licenses/by-nc/4.0/.

Ghana Corpus

Verse-aligned text for Ghanaian languages, plus several world languages, for building monolingual and parallel corpora. Every language is aligned on a shared verse key, so any single language can be pulled on its own or any two joined into a parallel corpus:

• —Monolingual corpus for any single language
• —Ghanaian ↔ English (English is the default pair)
• —Ghanaian ↔ Ghanaian (e.g. Twi ↔ Ewe)
• —Ghanaian ↔ other language (French, Spanish, Portuguese, German, Italian, Arabic, Chinese, Swahili)

How to use

This dataset is meant to be used through the Ghana Corpus Builder library, which downloads only the files you need and joins them for you:

👉 https://github.com/GhanaNLP/ghana-corpus-builder

pip install huggingface_hub
git clone https://github.com/GhanaNLP/ghana-corpus-builder.git
cd ghana-corpus-builder

# Twi ↔ English
python ghana_corpus.py --source twi

# Twi ↔ Ewe, a random 5,000-pair sample
python ghana_corpus.py --source twi --target ewe --limit 5000 --sample

# Monolingual Twi
python ghana_corpus.py --source twi --monolingual

# List every available language
python ghana_corpus.py --list

Each reference language has several Bible versions — a classic plus contemporary modern-language translations (English: CEB, ERV, CEV, GNT; French: LSG, BDS, Parole de Vie, Segond 21; Spanish: RVR1960, NVI, NTV, DHH; and the same for Portuguese, German, Italian, Arabic, Chinese, Swahili). By default all versions of a reference language are merged, so each Ghanaian verse is paired with every rendering (more paraphrases). Pin one version with @<id>:

python ghana_corpus.py --source twi --target en@406   # English ERV only
python ghana_corpus.py --source twi --target fr@21    # French Bible du Semeur only

You can also load the raw CSVs directly:

from huggingface_hub import hf_hub_download
import pandas as pd

path = hf_hub_download("ghananlpcommunity/ghana-corpus",
                       "Ewe_ewe_v1613.csv", repo_type="dataset")
df = pd.read_csv(path)

Layout

The dataset is split into three viewer configs (ghanaian, english, reference) because the file groups have different columns:

All files share the verse_key column (e.g. JHN.3.16), which is how any two languages are aligned. The filename itself encodes the language name, code, and Bible version id, so each file is self-describing.

Identifying a language from a row: the reference config also carries a lang_code column directly. For Ghanaian rows, the version_id column maps to a language via `youversion_ghana_versions.csv`.

Source & license

This dataset is released under CC BY-NC 4.0 — free to use, share, and adapt for non-commercial research and educational purposes with attribution.

Acknowledgements

Built by the Ghana NLP Community. If you use this data in research, please cite the community and acknowledge the underlying Bible-translation sources.