View all News

Google, University of Ghana and GDI Hub to Expand AI-Powered Speech Recognition for Non-Standard Speech in Ghanaian Languages

Accra, Ghana, November 1, 2024 – The University of Ghana, Google Research Africa, and UCL’s Global Disability Innovation Hub (GDI Hub) are collaborating on "tɛkyerɛma pa" ("Good Tongue"), an AI-based initiative aimed at improving communication for individuals with non-standard speech patterns. This project focuses on improving AI-powered speech recognition technology for five major Ghanaian languages, marking a significant step in making communication more inclusive for all.

On November 1, a tech talk and press event will be held at the Google Ghana office to present the progress of Project Euphonia, a research initiative designed to improve Automatic Speech Recognition (ASR) for individuals with non-standard speech. This event will highlight the upcoming collaboration between the University of Ghana, GDI Hub, and Google to expand the reach of AI-powered speech recognition to Akan, Ewe, Ikposo, Dagbani, and Dagaare.

Globally, many individuals with non-standard speech due to conditions like Cerebral Palsy, ALS, stroke, Down Syndrome, or Parkinson’s can face significant barriers in communication. Current speech recognition technologies often struggle to accurately interpret non-standard speech, limiting their effectiveness. The tɛkyerɛma pa project aims to address this gap by training AI models to better understand these speech patterns in local languages, ultimately enhancing accessibility and making speech recognition more accessible to Ghanaians with non-standard speech.

Project tɛkyerɛma pa is a collaborative effort led by the University of Ghana, supported by a $40,000 grant from Google and research expertise from UCL’s GDI Hub provided through the AT2030 program. Google Research scientists will directly train and support University of Ghana researchers and students to develop and refine speech models for the five languages. This initiative underscores the theme of "AI by Africa for Africa," showcasing how locally-led AI innovation can have a broader, global impact.

We are thrilled to lead this important initiative, which reflects our commitment to using technology to solve real-world challenges. By developing localised AI solutions, we can ensure that more Ghanaians have the opportunity to communicate effectively, regardless of their speech patterns."

Prof Isaac Wiafe, Project Lead / Associate Professor, University of Ghana

We are excited to collaborate with the University of Ghana on Project tɛkyerɛma pa, supporting their efforts to create the first open-source dataset of non-standard speech in Ghanaian languages. This project reflects Google’s ongoing commitment to supporting AI research in Africa and ensuring that technological advances benefit everyone, regardless of how they speak. Project ‘tɛtwrɛma pa’ builds upon continuous research in making Automatic Speech Recognition (ASR) more accessible. We hope to use the insights we’ve learned from our work in Project Euphonia to help local researchers create ASR models for local languages.

Katrin Tomanek, Senior Research Scientist, Google Research

Expanding the reach of AI-powered recognition to more languages is crucial to ensure people living with non-standard speech can communicate more effectively with people and devices in their daily lives. This collaboration will not only benefit individuals in Ghana but also provide valuable insights for global AI development. It will lay the foundation for future language models and collection in Africa, which we will support through our AT2030 programme.

Prof Cathy Holloway, Academic Director, GDI Hub

This initiative is part of Google's broader commitment to advancing AI development in Africa. By providing funding, technical expertise, and cloud computing resources, Google aims to foster an ecosystem where local researchers can lead innovations that address both local and global challenges. The project builds on the success of the Project Relate app, which continues to improve communication for individuals with non-standard speech worldwide.

The event on November 1 will mark the launch of the data collection phase, where local speech therapists will gather diverse speech samples from individuals across Ghana. Following this, the project team will train and refine ASR models tailored to the nuances of non-standard speech in Ghanaian languages. Looking ahead, the team aims to expand this technology to other languages and regions, ensuring the benefits of AI are shared widely.