Bridging the Digital Divide: New AI Platform Supports African Languages

In a significant step toward digital inclusion, African Languages Lab has launched Mansa, an AI platform offering multilingual and multimodal capabilities across more than 30 African languages. The initiative addresses a critical gap in natural language processing resources for Africa’s diverse linguistic landscape.

Addressing Data Scarcity

The foundation of this project stems from years of addressing the lack of quality data for training AI systems on African languages. Founder Sheriff Issaka explained that while building language technology solutions, they consistently encountered limited and often inadequate datasets. This led the company to proactively collect missing data, resulting in what’s now recognized as the largest collection of African-language data globally.

Key Features & Impact

  • The platform currently supports around 30 languages including Amharic, Hausa, Zulu, Twi, Igbo and Yoruba
  • It boasts over 100 billion curated tokens (text pieces) and 19,000 hours of validated speech recordings
  • Mansa is accessible via web, mobile apps, and APIs for developers

This initiative aims to address a fundamental challenge: Africa’s 2,000+ languages receive disproportionately low digital representation compared to their prevalence across the continent.

Addressing Cost & Safety Concerns

The team highlighted additional issues they’re tackling:

  • Cost inefficiency: Generating tokens in African languages can cost significantly more than in English for comparable performance
  • Safety gaps: Limited training data creates weaker safety filters, potentially exposing users to higher rates of inaccurate or inappropriate responses compared to well-resourced languages

Community-Driven Data Collection

A decade of dedicated effort has built this resource through:

  • Leveraging open datasets and community contributions
  • Direct engagement with language communities who validate and contribute new data
  • A platform called All Voices that enables direct data collection between African languages, bypassing the need for English as an intermediary

This approach empowers local communities to shape AI solutions that better serve their needs while addressing a critical digital equity issue.