By Shumirai Chimombe
“Language is the most important thing for the survival of Africans – and many other people in the world – but for Africans it has assumed a special significance. Africa is the second biggest land mass on Earth. It also has, population-wise, about 1.4 billion people. It is estimated that by 2050 almost half of the world’s population may be African. So what happens in Africa is bound to have a tremendous effect on the prospects of us all as a global community.”
This was part of a presentation delivered by Professor Kwesi Kwaa Prah, a sociology professor and the founder and director of the Centre for Advanced Studies of African Society, during a meeting hosted by the University of Pretoria in July last year.
In his talk, titled ‘African Languages in an AI World’, Prof Prah indicated that Artificial intelligence (AI) has revolutionised life in terms of using technology to acquire, accumulate and use knowledge. This is why it has become vital for African languages to be integrated into the development of AI technology to preserve cultural and linguistic heritage.
“Now what Africa faces in my estimation should be more work done on our languages and the preservation of our languages because that is the heart – the center of everything. Culture is the determinant of human beings and we have to make sure that African languages are written first. Without that we can’t make any progress.”
He referred to the ‘AI and the Future of Work in Africa’ whitepaper written by a cross-organisational team including Microsoft Research, Microsoft Philanthropies, University of Pretoria, NEPAD, Lelapa AI, and Oxford University. The whitepaper noted that while African languages are increasingly represented in AI large language models (LLMs) they lag behind English performance substantially, and currently only a small number are well-represented.
African culture and context are also notably underrepresented in generative AI training data, leading to poor performance in African workplaces. This is why it is essential to use representative African data to build models which work in African contexts. In addition to creating equitable data ecosystems, it is also important to incorporate indigenous knowledge in culturally and socially sensitive ways.
Africa.com echoed this sentiment, indicating that the inclusion of African languages in AI systems is a critical challenge that threatens to widen the digital divide, particularly in regions where local languages play a central role in daily communication. Developing language models that understand and generate text in languages like Swahili, Zulu, and Hausa is crucial for ensuring that Africa is not left behind in this technological era.
Responding to the call for AI inclusion
According to Africa.com African countries are now embarking on projects to integrate AI into local languages and develop LLMs that are tailored specifically to their diverse local languages. By doing this, African tech innovators are laying the foundation for a more inclusive digital future.
In South Africa for example, the company Lelapa AI has developed their VulaVula project, an AI-driven language processing tool that facilitates communication across several local languages, including Zulu, Sesotho, and Afrikaans.
The Nigerian government, in collaboration with local AI startups, is developing and training an LLM specifically to cater for the country’s very diverse linguistic community. This initiative is supported by a large network of volunteers fluent in Yoruba, Hausa, Igbo, Ibibio, and Pidgin, who are contributing to the data collection process.
Researchers from across the continent are collaborating on an open source AI project to develop machine translation for African languages. Masakhane – meaning “We Build Together” in isiZulu is a grassroots organisation whose mission is to strengthen and drive natural language processing (NLP) research in African languages, for Africans, by Africans.
According to Masakhane, “despite the fact that 2000 of the world’s languages are African, African languages are barely represented in technology. The tragic past of colonialism has been devastating for African languages in terms of their support, preservation and integration. This has resulted in a technological space that does not understand our names, our cultures, our places, our history.”
This cross-cutting collaboration to level the digital and AI playing field consists of about 1000 participants from 30 African countries with diverse educations and occupations, as well as about three countries outside Africa.
Overcoming the barriers
Although progress is being made in making AI more accessible, there are still some specific challenges in developing models in African languages. One of the most significant issues is related to the scarcity of data. Many African languages are considered low-resource, meaning that there is a limited amount of digitised text available to train AI models effectively. Also, in many African communities, oral traditions are the primary means of communication and exchanging information.This raises ethical questions about consent and ownership when collecting data.
By addressing these challenges around data scarcity and ethical concerns, African innovators are not only preserving their linguistic heritage but also paving the way for a more equitable digital landscape, concludes Africa.com in its article, ‘AI’s role in preserving and promoting African languages.’
Sources: University of Pretoria | Artificial Intelligence News | AI and the Future of Work in Africa | Africa.com | Masakhane
