Skip to content

Dataset · uvci

Speech and translation for the Dyula language

uvci/koumankan4dyula

A corpus of 15 hours of audio in West African Dyula with transcriptions and translations in French and English.

What's inside

The set includes 10,929 audio recordings (about 15 hours) in the low-resource Dyula language, spoken by over 16.4 million people in West Africa. Each recording is accompanied by source text and translations into French and English. Data collected as part of the Koumankan project.

What you can do with it

  • Training ASR models for speech recognition in the Dyula language
  • Training direct speech-to-text translation systems
  • Fine-tuning multilingual models on low-resource African languages
Tasks
Automatic speech recognition, Translation
Languages
Bambara
Size
10K–100K rows
Downloads this month
194
Likes
12
Access
on request on Hugging Face

How to load it

from datasets import load_dataset

ds = load_dataset("uvci/koumankan4dyula")