Hugging Face's logo

Datasets:
tonative
/
swahili-parallel-text-extension

Modalities:
Text
Formats:
parquet
Size:
10K - 100K
Libraries:
Datasets
pandas
Polars
Dataset card Data Studio Files
xet
 Community
1
Dataset Viewer
The dataset viewer is not available for this dataset.
Unexpected token '<', "<html> <h"... is not valid JSON

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

Swahili–English Extension of Kidaw’ida, Kalenjin and Dholuo Parallel Corpus

Dataset Description

This dataset is an extension of the parallel corpora introduced in Building Low-Resource African Language Corpora: A Case Study of Kidaw’ida, Kalenjin and Dholuo by:

  • Ambrose N. Ambogho
  • Quin Awuor
  • Andrew Kipkebut
  • Lilian Wanzare
  • Vivian Oloo

The original dataset provides Swahili–[Kidaw’ida/Kalenjin/Dholuo] parallel sentence pairs. English was not included in the original corpus.

Our Contribution

We extend the original dataset by:

  1. Using Machine Translation (MT) to translate the Swahili side of the dataset into English.
  2. Conducting human validation of the English translations with language speakers.
  3. Releasing both:
    • machine_translated_english
    • validated_english

This results in a new Swahili–English aligned dataset, while preserving the integrity of the original Swahili source sentences.

Please note that this data is still experimental and the validation may still be reviewed/improved.

Dataset Structure

Features

  • swahili (string): Original Swahili sentence from the source dataset
  • machine_translated_english (string): Raw MT output from Swahili → English
  • validated_english (string): Human-reviewed and corrected English translation

Splits

  • train
    • Examples: 29,230
    • Dataset size: 6,491,320 bytes

Motivation

English parallel data significantly improves:

  • Cross-lingual transfer learning
  • Multilingual foundation model training
  • Evaluation benchmarks
  • Accessibility for global research communities

By adding validated English translations, we increase interoperability of these low-resource African languages with broader NLP ecosystems.

Data Citation

If you use this dataset, please cite both:

  1. The original authors.
  2. This extended English translation version.

License

This dataset follows the original dataset license:

CC-BY-NC-4.0

Please review the original paper and dataset terms before commercial use.
Downloads last month
9
Total file size:
3.89 MB

Collection including tonative/swahili-parallel-text-extension