Language AI is changing the world as we know it. With natural language processing (NLP), natural language understanding (NLU), and the emerging field of natural language generation (NLG), we are able to communicate in different languages, ask our smartphones for directions, and talk with chatbots. As the technology continues to evolve and new products and features are released every day, it is obvious that we are undoubtedly in a golden age of language AI. This could not come at a better time, as AI in general is dominating cultural conversations as it continues to simplify our lives.
African startups are also creating exciting products with language AI models trained in African languages. Most NLP models are only trained on high-resource languages, meaning languages that have a large amount of data available for training AI models. African languages, on the other hand, are usually understudied, but African startups are now addressing the lack of data and resources, and creating models that cater to African languages while preserving and protecting them. Here are five startups that are changing the landscape of language AI in Africa.
VulaVula
One of the major hurdles with NLP in Africa is the absence of research on under-represented languages, like most African languages. Incubated within Lelepa.ai, a research and product lab that produces Africa-centric AI research, VulaVula is a language technology service that uses text and speech technologies to increase computational understanding of underrepresented languages, especially South African languages.
VulaVula, which means ‘speak’ in Xitsonga, features multilingual named entity recognition (NER), intent detection, and entity recognition as its key characteristics. With entity recognition, for example, the model is able to recognize pronouns with African origins. African names are recognized as persons and locations as locations. This helps to provide the context that might be missing when working with African languages.
GotBot
Based in Johannesburg, South Africa, Gotbot is a conversational AI chatbot provider that automates and improves the quality of responses to a business audience in one central messaging dashboard. These chatbots can be used on some of the most popular platforms like WhatsApp, LinkedIn, and Facebook, and can easily be integrated within a business to execute tasks including onboarding new clients and opening new accounts.
These chatbots not only automate daily repetitive tasks, they can also act as extensions of a business’s sales and support teams. Gotbot was founded in 2016 and has since worked with leading African brands to provide individualized customer service.
Botlhale AI
Speech based human-computer interactions are now commonplace in our everyday lives. Apart from their usual usage, we are using these programs to teach people how to read and pronounce words, read and transcribe for disabled people, and act as virtual assistants to millions of people. Unfortunately, most of these features are not accessible to people who only speak indigenous languages. Bolthale AI aims to provide an alternative for those people.
Through their wide range of products, including voice biometrics specifically for African speakers, voice enabled chatbots for African languages, and APIs that allow other people to build speech enabled solutions in African languages, they are eliminating language barriers for African users.
Mymanu
Traveling the world can be difficult if you are not able to interact with people who don’t speak your language. A Ghanaian-British entrepreneur founded Mymanu, a startup that aims to connect people through language. One of their products, Mymanu Clik S, uses language AI to enable real time translation through earbuds.
These earbuds enable travelers to speak to more than 2 billion people in 47 different languages around the globe. The earbuds also support voice assistance and live translation in local languages, making it easier to experience local life and explore hidden gems.
Lesan AI
With 22 endangered languages in Ethiopia alone, Ethiopian languages are some of the most understudied in the world. Because there isn’t a lot of data on them, Ethiopian languages are left out of many AI models and internet content in general. This makes it difficult for many Ethiopians to access the internet and interact with content on the web. Lesan AI is trying to change this.
Based in Berlin, Lesan AI is an automated translation service that has built translation tools for Amharic and Tigrinya, with plans to expand to other languages in Ethiopia. By creating these tools, they hope to fight misinformation on the internet and open up access to educational resources on the internet so that both children and adults can enjoy learning.
Conclusion
A lot of startups and organizations in Africa are working to ensure the representation of African languages in NLP. One of those organizations, Masakhane (which means “We build together” in isiZulu) intends to widen NLP participation in Africa and encourage NLP research for African languages. Other organizations like Mbaza-NLP, NLPGhana, and Deep Learning Indaba are also helping to build open source resources for NLP research.
Language AI in Africa is a rapidly expanding field, with many startups creating solutions that are accessible to African people in their own language. Because of the lack of data and research on African languages, a lot of work still needs to be done to ensure that these languages are represented in NLP and are considered when LLMs are being built. This would mean that African people are able to use language AI to communicate, interact, and engage with the world as they would elsewhere in the world.
Note: If you like this content and would like to learn more, click here! If you want to see a completely comprehensive AI Glossary, click here.