Improving speech recognition technology for AAVE (African American Vernacular English) speakers 

Status: Ongoing |  Funding: Howard University, Google

My research project focuses on improving speech recognition technology for African American Vernacular English (AAVE) speakers. We have collected over 200 hours of AAVE audio data, which will be used to train a new speech recognition model and evaluate existing services. The main goal is to address the bias in current speech recognition technologies that perform poorly for AAVE speakers by evaluating their performance, developing improved models using our AAVE dataset, and demonstrating the existence and extent of this bias. 

Contributors

Progress Timeline

• 2024-Present