Google AI researchers working with the ALS Remedy Improvement Institute right this moment shared particulars about Mission Euphonia, a speech-to-text transcription service for individuals with talking impairments. Additionally they say their method can enhance automated speech recognition for individuals with non-native English accents as effectively.
Folks with amyotrophic lateral sclerosis (ALS) usually have slurred speech, however current AI programs are usually educated on voice knowledge with none affliction or accent.
The brand new method is profitable primarily because of the introduction of small quantities of information that represents individuals with accents and ALS.
“We present that 71% of the advance comes from solely 5 minutes of coaching knowledge,” in keeping with a paper revealed on arXiv July 31 titled “Personalizing ASR for Dysarthric and Accented Speech with Restricted Information.”
Personalised fashions have been in a position to obtain 62% and 35% relative phrase error price (WER) enchancment for ALS and accents respectively.
The ALS speech knowledge set consists of 36 hours of audio from 67 individuals with ALS, working with the ALS Remedy Improvement Institute.
The non-native English speaker knowledge set known as L2 Arctic and has 20 recordings of utterances that final one hour every.
Mission Euphonia additionally makes use of strategies from Parrotron, an AI device for individuals with speech impediments launched in July, in addition to fine-tuning strategies.
Written by 12 coauthors, the work is being introduced at Worldwide Speech Communication Affiliation, or Interspeech 2019, which takes place September 15-19 in Graz, Austria.
“This paper’s method overcomes knowledge shortage by starting with a base mannequin educated on hundreds of hours of normal speech. It will get round sub-group heterogeneity by coaching customized fashions,” the paper reads.
The analysis, which a Google AI weblog submit highlighted right this moment, follows the introduction of Mission Euphonia and different initiatives in Could, equivalent to Stay Relay, a function to make telephone calls simpler for deaf individuals, and Mission Diva, an effort to make Google Assistant accessible for nonverbal individuals.
Google is soliciting knowledge from individuals with ALS to enhance its mannequin’s accuracy and is engaged on subsequent steps for Mission Euphonia, equivalent to utilizing phoneme errors to scale back phrase error charges.