Artificial intelligence (AI) research and product lab Lelapa AI wants to bolster its commercial offering locally by adding text-to-speech capabilities to its product line-up.
Furthermore, it is looking to expand its language base outside of Africa, as it sets out to build global language AI.
This is according to Pelonomi Moiloa, CEO and co-founder of Lelapa AI, in an interview with ITWeb TV, detailing the start-up’s founding roots and growth prospects.
During the wide-ranging interview, Moiloa discussed the company’s mission to build efficient AI for African languages; building smaller, data-efficient language models; how the local start-up ecosystem isn’t designed for small businesses to succeed; misunderstanding and fears about AI; as well as the greater awareness of AI, even among elderly people.
She also spoke about the need for humans to solve human problems, and how young people can position themselves to take advantage of the AI opportunities that lie ahead.
Multilingual tech tools
Despite AI language technology making significant strides, this technology revolution still marginalises African languages. Among the notable challenges is collecting data in African languages, given the costs involved and the quality of existing resources.
Upon identifying these challenges, Lelapa AI was founded in 2022, to determine how AI can be used for solutions and applications from an African lens.
The product lab currently has a live application programming interface (API) called Vulavula, which means “to talk” in xiTsonga.
The natural language model solution converts multilingual call centre conversations into structured text, for analytics, quality assurance and compliance systems. It is built for contact centres in telecoms and financial services.
Moiloa tells ITWeb TV that the API is for speech-to-text translation, so as one person speaks one language, the other language comes out in text.
“We want to add text-to-speech after that and have voice generation, so that you can plug it into anything, and have devices speak vernacular – that would be cool. We’ve done it for three languages now, and we would like to roll that out against all the languages that we offer.”
She added that Lelapa AI is excited about expanding its language base outside of the African continent. “We do believe that we built a technology that’s beneficial for a majority world, smaller models, less data, and that’s what we’ve set out to achieve.
“From a revenue perspective, we just want to grow, make money, sustain ourselves, make sure we can continue doing this mission and vision we’ve set out.
“At the end of the day, we want to be this bridge between digital products and services.”
An engineer turned data scientist, Moiloa, together with her co-founders, decided to focus on catering to the needs of a multilingual South Africa.
Lelapa AI’s other founding members include Jade Abbott and Pravesh Ranchod, and professors Vukosi Marivate, Benjamin Rosman and George Konidaris, who have backgrounds in academics, research, data science and engineering.
Moiloa explained that the start-up is focused on solutions for the enterprise space.
“They [enterprises] have large bodies of people that are communicating with them as an institution, but also them and their systems.
“Their internal systems are functioning in business language in the South African context, which is English. But there are Zulu, Xhosa, Swati, Sotho, Tsonga and Afrikaans-speaking people communicating with this institution that speaks English.
“We provide the language infrastructure layer between the people and that institution, so that they can communicate with each other. Within that institution, it’s not just the people; it’s also the systems that they build.”
Lelapa AI’s role is to take the “messy” data of the speech world and of the multilingual world and help people structure it into a way that is then searchable and can be analysed, she stated.
“What that means, for example, in a call centre you have people calling in. The agent is multilingual, but then all the audio recorded is in a local language that the English institutional system can’t look at, can’t understand, can’t search, can’t analyse.
“We’ll transcribe that, translate it into English, so that it fits within their analytic systems. It looks like having a chatbot, or a WhatsApp bot where people can engage in vernacular, allowing for the communications that go out to be in a relevant language for somebody to communicate with.”
Encouraging signs
Through its machine translation models, Lelapa AI has used 60% less compute and 60% less data than what is normally required to “just tweak” a model to build a language.
This, she said, showcases opportunities to the rest of the world that creating technologies for different languages doesn’t have to be tricky.
Moiloa revealed that since going live with Lelapa AI four years ago: “We have seen that the models have gotten better on certain local languages where they have access to data or large bodies of data that they can train models. Even on the hyperscaler side, we’ve seen a bit of higher literacy in terms of AI.
“When we first started Lelapa, we never use the word AI – we were always talking about machine learning, and that has since changed.
“There really is a greater consciousness around the technology itself, and things have moved, and they move so fast. What we tend to find is that the true value isn’t necessarily in the models themselves, but where they provide value, and that has been the tricky part that hasn’t changed much.
“People still don’t fully understand how to turn this cool, amazing thing into things that increase revenue, that lower their opex lines.”
For young people looking to position themselves to take advantage of AI opportunities, she said it’s critical to view AI as a guiding implementer of something they have come up with, rather than depending on it to produce a solution, because it’s just going to produce something that it’s seen before.
“Getting familiar with the tools, knowing how to utilise them towards your will and your ideas and your implementations is critical. At the end of the day, we still need humans to solve human problems.”
Share
