Google, in collaboration with Howard University, unveils a new dataset aimed at enhancing AI's comprehension of African American English.
In a groundbreaking initiative, Project Elevate Black Voices, a collaboration between Howard University and Google Research, has released a comprehensive dataset of over 600 hours of African American English (AAE) dialects from 32 states. This project aims to address long-standing issues of inaccurate AI results when interacting with Black users, particularly in voice assistant technology.
The dataset was meticulously collected through grassroots data collection, with the project team traveling across the United States to record speech at community events. This approach ensured the capture of diverse dialects and diction commonly used in Black communities but often overlooked by AI-driven automatic speech recognition (ASR) technologies.
The dataset, owned by Howard University, is being licensed to allow Google to use it for product enhancement. However, the university is taking great care to ensure the data is used ethically, prioritizing its use by those who align with values of inclusivity and cultural respect.
Dr. Gloria Washington, a Howard University researcher and co-principal investigator of Project Elevate Black Voices, emphasized the historical significance of African American English (AAE), a dialect that has been at the forefront of United States culture since almost the beginning of the country.
The goal of the project is not limited to African Americans but extends to individuals who speak unique African American English dialects. By enhancing AI's understanding and recognition of these dialects, the project aims to improve the way Black people interact with technology, making it more inclusive and accessible.
The issue of code-switching in ASR-based technology has been a significant concern, with Black users often altering their voices to be understood. This dataset, with its focus on diverse AAE dialects, is a step towards addressing this issue, ensuring that AI technology better reflects and caters to the diverse linguistic landscape of Black communities.
References: [1] Google Research. (2021). Project Elevate Black Voices. Retrieved from https://research.google/projects/project-elevate-black-voices/ [2] Howard University. (2021). Project Elevate Black Voices. Retrieved from https://www.howard.edu/research/project-elevate-black-voices [4] Washington, G. (2021). The Importance of Project Elevate Black Voices. Retrieved from https://medium.com/google-research/the-importance-of-project-elevate-black-voices-290460f5b4c3
This initiative, a collaboration between Howard University and Google Research named Project Elevate Black Voices, endeavors to improve artificial-intelligence (AI) technology across education-and-self-development platforms and personal-growth, focusing on the understanding and recognition of unique African American English (AAE) dialects. This project, in alignment with values of inclusivity and cultural respect, aims to reduce the need for code-switching in AI-driven automatic speech recognition (ASR) technology, making technology more accessible for diverse Black communities.
