Project Grant 2202467

This $178,992 National Science Foundation project grant supports research at Georgia State University Research Foundation to improve speech technology for better learning outcomes among African American English-speaking children. Funded under the Computer and Information Science and Engineering program, the three-year award aims to develop data augmentation and normalization techniques to increase the amount of training data available and improve automatic speech recognition performance for this underserved population. Researchers will utilize a developing database of African American English children's speech to examine the impact of spoken language systems on literacy assessment outcomes. The work also explores new methods for text-to-speech to model dialect density and account separately for pronunciation and prosodic factors. Outcomes include algorithms for low-resource tasks and leveraging effective text-to-speech methods in data augmentation for both automatic speech recognition and literacy assessment diagnostics.