Google And Howard University Collaborate To Help Devices Better Understand Black People’s Voices
Google and Howard University have announced a new partnership, Project Elevate Black Voices, to build a high-quality African-American English (AAE) speech dataset.
The new partnership will allow Howard University to share the dataset with those looking to improve speech technology while establishing a framework for responsible data collection, ensuring the data benefits Black communities.
Google has confirmed that the university will retain ownership of the dataset and licensing.
Automatic Speech Recognition and Black People
Automatic Speech Recognition (ASR) technology allows humans to use their voices to speak with a computer interface.
Oxford Academic reported on several studies that have found that widely-used ASR systems function much more poorly on the speech of Black people.
A study on ASR performance with racialized varieties found that ASRs from Google, Apple, Amazon, and Microsoft performed significantly worse on the speech of Black Americans than white Americans.
Other studies found that widely tapped speech corpora – a database of speech audio files – used to develop and evaluate speech recognition systems displayed a lack of representation of African American language.
The biases have begun to display adverse effects on African American speakers, with others finding that ASR failures hindered participants in accomplishing goals.
It also caused them to experience emotions consistent with those experienced during discrimination in human interaction.
Google’s research also found that Black people in the US often have a worse experience using ASR when compared to white speakers.
Due to this, Black users changed their voice patterns to be understood, known as code-switching.
Project Elevate Black Voices
The program Project Elevate Black Voices aims to make it easier for Black people to use ASR, Google stated.
Google’s responsibility will include performing user experience research to collect input from the community, providing infrastructure for data collection, funding and obtaining a license to use the data commercially.
Howard University will lead the coordination effort with other HBCUs, manage vendors, recruit participants and maintain the repository.
“There’s a lot of distrust and mistrust, rightfully so, towards technology. We want to make sure that when we’re collecting something as sensitive as voice data, which is considered biometric data,” Google Responsible AI Research and Social Psychologist Dr Courtney Heldreth told The Root.
“We are doing it in partnership with folks who are very connected to and understand the Black community,”
Project EBV’s Principal Investigator and Howard University Associate Professor Dr. Gloria Washington added, “The reason this project is so important – from my perspective – is that by gathering audio data to help voice assistant technology, we’re allowing Black people to tell their stores.”