N-ATLAS, an open-source, multilingual, multimodal large language model that supports Yoruba, Hausa, Igbo, and Nigerian-accented English, was unveiled by the Federal Government of Nigeria at the 80th United Nations General Assembly in New York.
In a statement posted on his X account on Saturday, Dr Bosun Tijani, Minister of Communications, Innovation, and Digital Economy, said N-ATLAS was created with African perspectives at the forefront of AI development.
He said, “Starting with Yoruba, Hausa, Igbo and Nigerian-accented English, N-ATLAS places Africa’s voices and diversity at the foundation of AI.
“This is the first step in a broader journey to make Africa a contributor and leader in shaping AI’s future.”
Functionality of N-ATLAS
According to the product description on its model card, N-ATLaS LLM has a speech-technology suite with language-specific automatic speech recognition models supporting local-language applications, accessibility, and transcription.
It can be used to create chatbots that respond to public inquiries in Hausa, Igbo, Yoruba, or English with a Nigerian accent.
The ASR models can produce captions and subtitles in the four local languages and convert radio, television, or internet videos into text.
Voice assistant and call centre support can use ASR and LLM pipelines to record caller speech in Nigerian accents, transcribe it, and provide automated responses or intent detection.
Additionally, interviews conducted in local languages can be summarised using the model.
The product was developed by the National Centre for Artificial Intelligence and Robotics and collaborators, including Awarri Technologies, as part of Nigeria’s language-AI initiative.
The model documentation can be viewed on Hugging Face.