The Minister for Communication, Digital Technology and Innovations, Samuel Nartey George, has called for greater use of Ghanaian indigenous-language datasets to accelerate the development of artificial intelligence tools that understand local languages and communication patterns.
He said the availability of approximately 35,000 hours of Ghanaian indigenous-language data presents an opportunity for researchers to build AI models that respond more effectively to the needs of local users.
Mr George made the remarks during discussions on language technology research involving Ama Branoa Banful, a PhD student at the Kwame Nkrumah University of Science and Technology (KNUST), who is working to develop speech recognition models capable of understanding English-Twi code-switching.
Ama Branoa Banful is seeking to address a major limitation in existing speech recognition systems, which often struggle to understand speakers who combine English and indigenous languages within the same sentence.
She explained that code-switching is a common feature of everyday communication in Ghana, where speakers frequently move between English and languages such as Twi, Fante and Pidgin.
The researcher said her doctoral work would build on research she began during her master’s programme, with the goal of developing a speech-to-speech pipeline that can process multilingual conversations.
“For my master’s, I started looking into how I can incorporate code-switching, which is when you mix two or more languages in a given sentence,” she said.
She noted that most existing models are trained on one language at a time, making it difficult for them to accurately interpret sentences containing multiple languages.
Ama Banful said the proposed technology could improve access to digital services, especially for people in rural communities who may find speaking easier than reading or writing.
“If you want someone in the village to benefit from what we are doing, we have to start getting into speech systems instead of just speech-to-text systems,” she stated.
Responding to the presentation, Sam George highlighted research collaborations involving Google and the University of Ghana, which supported the development of approximately 5,000 hours of data covering seven indigenous Ghanaian languages.
He explained that the datasets were intended for public use and could be shared with researchers at other institutions to support the development of language technology.
The minister also referenced a broader collection of approximately 35,000 hours of Ghanaian indigenous-language data, describing it as useful material for training AI models.
He encouraged Banful to build on the existing work and explore opportunities to develop solutions that could serve users beyond Ghana.
Hon. Sam George further pointed to research at the University of Cape Coast, including work by Professor Stephen Moore on languages such as Swahili, as an example of how local research could contribute to wider African technology development.
He stressed the importance of addressing not only Ghanaian challenges but also language-related barriers across the continent, where multilingual communication remains widespread.
Banful’s research could contribute to that effort by helping AI systems understand how people naturally speak, rather than requiring them to communicate in a single language.
The initiative forms part of broader efforts to develop locally relevant AI technologies capable of improving access to digital services and expanding the participation of indigenous-language speakers in the digital economy.
