The Koumankan4Dyula corpus consists of approximately 15 hours i.e. 10,929 recordings of Dioula language audio along with the corresponding texts and their translations into French and English. Dioula is a low-resource language spoken by over 16.4 million people in several West African countries. This corpus is part of the Koumankan project, which offers a scalable and cost-effective method to expand the CommonVoice dataset to include the Dyula language and other African languages. It will serve as a benchmark for training models for automatic speech translation in Dioula as well as for automatic speech recognition or machine translation models from Dioula to French and English.

Data Splits

split	percentage	Nm. records	Num hour
Train	73%	8065	8h 9m
Valid	14%	1471	1h 36m
Test	13%	1393	44m 36s

Maintenance

This dataset is supposed to be actively maintained.

Benchmarks:

Coming soon

License

CC-BY-SA-4.0

Version

1.0.0

Acknowledgements

This dataset collection efforts have been supported by International Development Research Centre (IDRC) and Swedish International Development Cooperation Agency (SIDA), managed by African Center for Technology Studies (ACTS) in collaboration with the Université Virtuelle de Côte d'Ivoire (UVCI) through the programme Artificial Intelligence for Development (AI4D) Africa.

Contacts

ismael21.kone@uvci.edu.ci
info@data354.com

Downloads last month: 379

Total file size:

4.93 GB

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

Koumankan4Dyula: A Speech-To-Text Translation Corpus For Dyula Language

Overview