a day ago
Enhancing research accessibility: University of Pretoria's innovative AI translation initiative
The University of Pretoria's library services has launched an AI project aimed at translating academic research into several languages.
Image: File
The University of Pretoria's Department of Library Services (DLS) launched the groundbreaking 'Translation of Abstracts' project - an initiative that harnesses artificial intelligence (AI) to make research accessible in multiple indigenous languages.
The project is the product of collaboration between professional and academic departments, reflecting the university's commitment to transdisciplinary work.
'This is about the kind of university we want to be,' said Professor Suni Maharaj, UP Vice Principal for Research, Innovation and Postgraduate Education, in stressing the need for collaboration and to bridge not only the digital divide, but also the emerging AI divide.
'AI is a tool that allows us to do things we never thought possible. We must ensure it becomes an opportunity for all, not a barrier for many.'
The DLS sees this project as more than just a technical exercise - it is an invitation for students to engage with issues beyond their academic timetables, through activities such as book clubs and poetry nights in indigenous languages, said the DLS Director, Lindiwe Soyizwapi.
It was further said that providing access to knowledge is central to the library's mandate, and with research capable of addressing societal challenges and shaping policy, this project aligns with a broader movement toward inclusive, multilingual knowledge that benefits both academia and society.
The project was linked to the long-held dream of a decolonised education system, in which students can think, learn, and create in their mother tongues. 'This is cultural and linguistic preservation. It's equity. It's self-determination. We are making history by ensuring no student is excluded,' student David Kabwa said.
A representative from the South African Centre for Digital Language Resources (SADiLaR), Professor Langa Khumalo, reinforced this by stating that language humanises us. 'The choice of language in education has historically been weaponised to suppress other ways of learning. We must actively cultivate indigenous digital knowledge, challenging the notion that English equals knowledge.'
Professor of Computer Science, Vukosi Marivate, explained that systems don't know our languages. 'We need to build systems that recognise them.' He added that this work is driven not by ego, but by care - care for communities, for future generations, and for the survival of linguistic diversity.
The project's development involved training AI models with 25 academic abstracts into Afrikaans, IsiZulu, and Sepedi, refining them through human translation and validation to ensure academic precision, accuracy, and language context.
Khumalo described the aim as 'not to dumb it down, but to maintain scientific rigour while broadening access.'
Early results have shown the unique challenges of different languages. For example, Afrikaans translations required only minor human editing, while Sepedi translations demanded significant reworking due to missing terminology - a process likened to 'building a house while manufacturing the bricks.'
