Wannan littafin yana da matukar amfani! Ina ba da shawara ga kowa ya karanta.
"This book is genuinely useful — I recommend it to everyone."
Masakhane Playbook
A practical guide to dataset creation, written with the communities who use it — from task formulation and label schema design to consent forms, inter-annotator agreement, and sustainability. Every chapter is built around real low-resource language scenarios.
- Step-by-step guidelines, video demos, and quality checklists
- Voice, text, speech–text alignment, and translation chapters
- Templates for consent, licensing, and governance toolkits
- Translated into 5 African languages with community review
Masakhane Tool
An open, mobile-first, Progressive Web App for grassroots data collection — built for the realities of African contexts: patchy connectivity, multiple scripts, and community-led annotation workflows.
- Offline-first capture with background synchronization
- Speech, text, ranking, and multimodal annotation support
- Inter-annotator agreement (Fleiss' κ, Krippendorff's α) dashboards
- African-language localization and virtual keyboards
- Apache 2.0 licensed with a clear contributor agreement
From the Community
The Playbook is exactly the practical, reproducible guide that African NLP has needed — a real reference, not a brochure.
Prof. Vukosi MarivateCo-founder, Masakhane · University of Pretoria
Pairing the Playbook with the Tool turns annotation theory into reproducible practice — that combination is what makes it useful in the field.
Dr. David AdelaniNLP Researcher, Masakhane · University College London
Documenting low-resource language work has long been ad-hoc — a shared playbook gives our teams a common vocabulary and saves a lot of guesswork.
Lilian WanzareNLP Researcher · Maseno University
Open guidance like this lowers the barrier for builders across the continent to ship language-first AI products responsibly.
Pelonomi MoiloaCo-founder & CEO · Lelapa AI
Open infrastructure for African languages is finally catching up with the rest of the field. This is a major milestone for the community.
GremaAI Researcher · Microsoft
A community-built standard for low-resource annotation. Long overdue and well executed — the kind of resource teams will reach for daily.
AishwaryaResearch, Language Technologies · Google
For multilingual annotation across Bantu languages, this is the resource I wish we had had years ago — clear, applicable, and honest about trade-offs.
Peter NabendeNLP Researcher · Makerere University
The combination of methodological rigor and African-context grounding makes this stand out from generic NLP guides — a long-overdue reference.
Prof. Muhammad Abdul-MageedNLP Lab Lead · University of British Columbia
Help build the Playbook
We welcome researchers, practitioners, students, and language experts. Pick whichever fits — every contribution counts.
Write a chapter
Fill a gap in the Playbook — propose a chapter, write it, open a PR.
Translate a page
Adapt an existing chapter into Hausa, Amharic, Swahili, French, or Portuguese.
Join Discord
Discuss approaches, share your work, and meet other contributors.
Star on GitHub
Help the project gain visibility — every star widens its reach.
Thanks to our Contributors
The Playbook is built by a growing community of researchers, students, and language experts. If you've contributed code, content, or review — thank you.
