Google is collecting voices of people with Down syndrome to develop more accurate voice recognition technology for those with the disorder.
The initiative, called “Project Understood,” is a collaboration between the tech giant and the Canadian Down Syndrome Society.
“My daughter has a smaller mouth and a larger tongue. Picture talking with a marshmallow in your mouth,” said Ed Casasgrande, Chair of the Canadian Down Syndrome Society, in a promotional video for the project.
Voice recognition technologies tend to miss every third word spoken by people with Down syndrome, “which makes the technology not very usable,” Google Engineer Jimmy Tobin said in the video.
“Because of their unique speech patterns, voice technology doesn’t always understand people with Down syndrome,” the project’s website says. “Project Understood is ensuring the future of voice technology includes people with Down syndrome.”
Speech patterns for people with Down Syndrome are unique and difficult for voice technologies like Siri and Alexa to detect because of muscular and skeletal structures — as well as “a large lack of training data,” the website says.
Bob MacDonald, technical program manager at Google, said that “technologies that are activated by voice commands are becoming a way of life.”
“For people where that doesn’t work, that must feel very disempowering or like they’re being sort of left behind,” MacDonald said.
The initiative hopes to collect 500 voices to train Google algorithms — and so far has collected 300.
“My friends can understand me. Why can’t voice technology?” project participant Joshua said.