Artificial intelligence is becoming part of everyday life, but for many Black users, getting technology to understand them can still mean changing the way they speak. Howard University and Google Research are working to change that with a project designed to make voice-recognition technology more inclusive.
What Exactly Is Project Elevate Black Voices?
Through Project Elevate Black Voices, researchers developed a dataset containing more than 600 hours of African American English (AAE) speech collected from participants across 32 states. The resource helps developers improve automatic speech recognition systems that power voice assistants, transcription tools, and other AI-driven
products. Speech-recognition systems can struggle with dialects and language patterns that are underrepresented in their training data. For Black users, those shortcomings can lead to misinterpreted commands and inaccurate transcripts, creating barriers to using tools that others may take for granted.“Voice assistant technology should understand different dialects of all African American English,” said Dr. Gloria Washington, a Howard University researcher and co-principal investigator of the project, in a press release. She added that doing so would help technology better serve African Americans and other speakers of these dialects.
The implications extend beyond convenience. As businesses incorporate
AI into customer service, workplace communications, and digital products, speech-recognition accuracy can affect how people access services and interact with technology. Improving the systems could help companies develop products that work for a broader range of customers.Project Elevate Black Voices also places community participation at the center of its research. The team organized events in multiple cities where Black community members discussed their experiences with technology, AI, and data privacy. Participants were then invited to contribute audio recordings during a three-week collection initiative.
Who Will Take Ownership Of The Data?
Lucretia Williams, the project’s lead and a Howard researcher, said the team wanted to create trusted spaces where participants could share their experiences.
“As a community-based researcher, I wanted to carefully curate the community activations to be a safe and trusted space for members of the community to share their experiences about tech and AI,” Williams said.
Howard University will retain ownership of the dataset and oversee its licensing and responsible use. Initially, the Howard African American English Dataset 1.0 will be available exclusively to researchers and institutions within historically Black colleges and universities, with broader distribution to be considered later. Google can also use the dataset to improve its products. Courtney Heldreth, a Google Research co-principal investigator, said the collaboration could help more people communicate naturally with smart devices.
The project reflects a broader push to address bias in AI development by ensuring that communities historically overlooked in technology research help shape the tools they use.
RELATED CONTENT: Howard University Appoints Carole Boyce-Davies, 1st Endowed Chair In Caribbean Studies