UNICEF WCARO NLP Landscape

West and Central Africa Language Technology
Last updated: 2026-04-23
Work in progress. Some data may contain inaccuracies. Contribute on GitHub

Jokalante

Startup established

Organization Information

Startup
2016
small
established
open
Dakar, Senegal
Lacuna Fund (KALLAAMA project, 2023). Fund for Innovation in Development / FID (NAFOORE chatbot). GSMA AgriTech Accelerator (via GIZ/BMZ). WFP Innovation Accelerator.
jokalante

Coverage

Senegal, The Gambia
WoloffucsrrdyoMandinka

Description

Senegalese social enterprise ("jokalante" = "facilitate dialogue" in Wolof) providing digital solutions for rural and underserved populations. Core platform provides IVR, SMS, voice, and data collection services across all 14 regions of Senegal, reaching 1M+ people in 7 languages (Wolof, Pulaar, Sereer, Diola, Mandinka, French, English). Primary focus is agriculture, with 250,000 direct beneficiaries. Created the KALLAAMA speech dataset (125 hours of transcribed agricultural speech in Wolof, Pulaar, and Sereer) funded by Lacuna Fund. Developing the NAFOORE AI chatbot providing personalized farming advice via WhatsApp and voice calls, adapting to weather, soil, and crop conditions using ANACIM meteorological data and CGIAR agroecological knowledge.

🦄 UNICEF Relevance

Directly relevant as a Senegal-based organization building speech technology and AI chatbots for local languages with agricultural and climate focus. KALLAAMA dataset covers Wolof, Pulaar, and Sereer -- languages that currently lack dedicated actors. The NAFOORE platform and IVR/SMS infrastructure reaching 1M+ rural users is exactly the kind of community engagement infrastructure UNICEF programs use. Relevant for the March 2026 Senegal summit.

Key People

Projects

Publications

Partnerships

Notes

Primary expertise is rural digital services, not AI research. Their entry into NLP is driven by practical need (reaching farmers who speak local languages). The KALLAAMA dataset is one of very few open speech resources for Pulaar and Sereer. HuggingFace profile has no public models or datasets yet (data hosted on GitHub/OpenSLR instead). 70% of surveyed users reported reduced material losses from timely weather alerts.

Last updated: 2026-01-26