Google Partners African Universities to Launch WAXAL, a Major Dataset for African Languages

Google has partnered with leading African universities and research institutions to launch WAXAL, a large-scale open speech dataset designed to improve artificial intelligence tools for African languages.

The initiative is aimed at closing a long-standing gap in voice-based technologies by providing high-quality speech data for languages that have historically been excluded from global AI systems. Through WAXAL, Google and its partners are laying the foundation for more inclusive, accessible, and culturally relevant digital tools across the continent.

Addressing Africa’s Language Gap in Artificial Intelligence

Voice assistants, speech-to-text systems, and automated translation tools have become essential in many parts of the world. However, Africa’s linguistic diversity has remained largely under-represented in these technologies.

With more than 2,000 languages spoken across the continent, only a small fraction have been properly documented in AI training datasets. This lack of data has made it difficult for developers to build reliable speech-powered tools for African users.

As a result, millions of people have been unable to fully benefit from voice-enabled services in education, healthcare, business, and public services. WAXAL was created to directly confront this digital inequality.

What the WAXAL Dataset Contains

The WAXAL dataset includes speech data from 21 Sub-Saharan African languages, including Hausa, Yoruba, Igbo, Luganda, Swahili, and Acholi.

According to Google, the dataset was designed to support more than 100 million speakers who have been left out of mainstream voice technologies due to limited language resources.

Key components of the dataset include:

• Over 1,250 hours of transcribed natural speech

• More than 20 hours of high-quality studio recordings

• Data suitable for training speech recognition and synthetic voice systems

• Materials that can be used for academic research and commercial development

These resources make it possible for developers to create more accurate, realistic, and culturally appropriate speech applications.

A Three-Year Collaborative Development Process

The WAXAL project was developed over three years with financial and technical support from Google. Rather than building the dataset independently, Google worked closely with African institutions to ensure local participation and ownership.

Universities and organisations such as Makerere University in Uganda, the University of Ghana, and Digital Umuganda in Rwanda led the data collection process. Local researchers and community members were actively involved in gathering, reviewing, and validating speech samples.

This collaborative model helped ensure that the dataset reflects real-life speech patterns, accents, and contexts rather than artificial or externally imposed standards.

Empowering African Innovators and Researchers

Speaking on the project’s long-term impact, Aisha Walcott-Bryantt, Head of Google Research Africa, emphasised its role in empowering local communities.

She explained that WAXAL provides students, researchers, and entrepreneurs with the foundation they need to build technology in their own languages and on their own terms. By doing so, it opens opportunities for innovation that directly responds to African realities.

This approach shifts Africa’s role in AI development from being mainly a consumer of foreign technologies to becoming an active creator of digital solutions.

Community Ownership and Data Control

One of the most distinctive features of WAXAL is its ownership structure. Unlike many global datasets that are controlled by multinational corporations, the partner institutions retain ownership of the data.

This ensures that African researchers, universities, and startups have long-term access and decision-making power over how the data is used. It also reduces dependency on external companies for core technological resources.

Joyce Nakatumba-Nabende, a senior lecturer at Makerere University, highlighted the importance of this model. She noted that for AI to truly serve Africa, it must understand local languages and contexts. According to her, WAXAL gives researchers the tools they need to develop technologies that reflect their communities.

Large-Scale Volunteer Participation

The success of WAXAL also depended heavily on public participation. At the University of Ghana alone, more than 7,000 volunteers contributed their voices to the project.

Professor Isaac Wiafe, an associate professor at the University of Ghana, described the initiative as a major driver of innovation. He explained that the dataset is already supporting new ideas in sectors such as healthcare, education, and agriculture.

By involving everyday speakers, the project ensured that the data represents diverse age groups, regions, and speech styles.

Expanding Opportunities for Inclusive Technology

With the dataset now publicly available, developers, startups, and academic institutions can access foundational speech data to build new products and services.

Potential applications include the following:

• Voice assistants in local languages

• Speech-based learning platforms

• Medical transcription tools

• Agricultural advisory systems

• Customer service automation

• Accessibility tools for people with disabilities

These innovations can help bridge communication barriers and improve digital participation across different communities.