University of Pretoria Pioneers AI Solutions for Indigenous Language Education

The University of Pretoria is spearheading a significant initiative that focuses on developing artificial intelligence (AI) technologies catering specifically to indigenous African languages. This approach aims to tackle existing barriers in the educational landscape, such as language discrepancies, a deficit of training data for AI applications, and particularly, the issue of overcrowded classrooms in South Africa. By creating language-specific datasets and tailored AI tools, the University seeks to foster better literacy rates and enhance educational outcomes across the continent, providing culturally relevant and inclusive learning experiences.

This undertaking aligns with the pressing demands of the current educational environment in South Africa, which is facing a literacy crisis. Many Grade 4 learners are unable to read proficiently in any language, as highlighted by Dr. Olaperi Okuboyejo, a Senior Lecturer in Computer Science at the University. According to the 2021 Progress in International Reading Literacy Study, approximately 81% of Grade 4 students struggle with reading comprehension. Okuboyejo attributes these challenges to the abrupt transition from mother-tongue instruction to English, a language that does not serve as the primary home language for most students, with only 7% speaking English at home. Furthermore, the performance of South African students in subjects such as mathematics and science continues to rank poorly on a global scale.

Overcrowded classrooms exacerbate these educational challenges, with an average learner-to-teacher ratio of 33:1, significantly exceeding the recommended ratio of 16:1. Okuboyejo argues that integrating indigenous language instruction is paramount to improving literacy, reinforcing cultural identity, and boosting self-esteem among learners. She stated, “Learning to read in the mother tongue is instrumental in unlocking every child’s potential,” emphasizing the need for accurately developed resources and terminology that support subjects like mathematics and science. In this context, AI offers promising solutions that can enhance personalized learning experiences and provide necessary tools for creating accessible educational content.

Despite promising technologies, researchers face a formidable barrier: the limited availability of language data. Okuboyejo pointed out the dire lack of online resources for languages such as Zulu, which impedes the development of effective language-specific models. To address this, the University of Pretoria is actively building extensive datasets encompassing over 3,000 hours of speech in various South African languages. The institution is also developing terminology lists for technical subjects and working on audio models for 40 different languages, along with translation projects aimed at facilitating academic research.

The call for collaborative efforts among key stakeholders such as governments, educators, media agencies, and researchers is crucial. Governments need to establish policies that encourage data utilization, educators must incorporate technology into the classroom actively, while media outlets should champion the promotion of indigenous languages. Okuboyejo aptly noted, “Closing the language gap in educational AI is central to closing the access gap in African education,” which advocates for a deliberate, not an accidental, integration of AI solutions into educational frameworks.

As the University of Pretoria continues to innovate, their work creates a pathway for future contracts and initiatives aimed at enhancing educational outcomes through technological solutions. Vendors specializing in AI, language processing, and educational technology will find opportunities to engage with this emerging market while contributing to government-supported projects designed to support these educational strategies. In summary, this initiative illustrates a significant shift in procurement requirements, potentially influencing future contracts towards cultural sensitivity and customized education technologies.