A corpus collected in the field by a community of linguists and native annotators, designed to cover real domains: health, administration, everyday conversation.
A corpus built by the community
Eyaa-Tom “words of the people” was born from a simple principle: a language cannot be digitized from an office, only with those who speak it. Linguists, researchers and native annotators collected the texts in the field — in markets, clinics and administrations.
Domains that truly matter
- Health: symptoms, advice, exchanges with healthcare staff
- Administration: paperwork, forms, official documents
- Everyday conversation: daily life and community topics
Quality and verification
Every sentence is reviewed by a second speaker before being integrated, and disagreements between annotators are settled collectively. This double validation ensures data that can be used to train and evaluate models.
Open release
The corpus is published on Hugging Face and can be reused by the research community and developers. It serves both for training Yodi models and for independent benchmarks on Togolese languages.