Fact · Accessibility and health
430speakers
Google trained personalized recognizers for about 430 speakers with speech impairments, each of whom had recorded at least 300 utterances.
Checked against its source on October 16, 2026
In the source's own words
“We trained personalized ASR models for ~430 speakers who recorded at least 300 utterances”
Personalized ASR Models from a Large and Diverse Disordered Speech Dataset, Google Research Blog, 2021, Results.
Open the original- What it measures
- Each speaker's personal model was tested on 10 percent of its utterances held out from training, with no phrase overlap.
- What it does not show
- A company blog summary. Speakers read prompts in a browser-based recording tool, so the recordings are not clinical recordings.