Advances in automatic voice recognition technology have spurred a Paris-based start-up to launch a crowd-fund campaign to bring to market the first worldwide service that offers captioned telephone calls for deaf people.
For the past two years, RogerVoice has been working on developing a subscription-based app that combines a Voice-over-IP telephone platform with streaming automated speech recognition to transcribe voice calls onto a smartphone, tablet or PC screen in real-time so that users who are deaf or hard of hearing can engage easily in a telephone conversation without relying on any human intervention.
The service works by starting a phone call to a landline or mobile number via Skype or any other VoIP service. "When your contact picks up the call, she receives a spoken message saying: 'This call is being transcribed . . . ' but aside from that, nothing distinguishes it from any other call," said RogerVoice founder Olivier Jeannal. "He or she speaks, and you receive a text transcription on your screen, in real time, during the conversation."
“RogerVoice counts most when you have no option but to call. Our app is designed to provide deaf people with as near to an instant and conversational phone experience as possible.”
But what about deaf users who cannot or don’t want to use their voice? Jeannal says he is working to get “text-to-speech synthesis” (TTS) in the first version of the app, which means that words a user types into the app can be relayed by the service using an automated voice facility, but this may not happen until early next year.
“Our first impetus was to get the hardest part working and get it out there: streaming speech-to-text transcriptions of phone calls. The second component, TTS, will be a stretch goal.”
The Kickstarter campaign, which was launched last week to coincide with International Week of the Deaf, aims to raise $20,000 to build a more solid and stable platform, which will initially be based on Android.
“The solution appears simple but the technology is incredibly hard to implement,” said Jeannal. “Our goal is to make it so that the end client doesn’t have to do any special tinkering.”
Automated voice recognition technology has come a long way, but is it accurate enough to work without any human intervention?
“Machines will never replace human ability to detect and understand speech,” said Jeannal. “In a world where the phone is still very much present, though, I figure voice recognition can help cope with 80 per cent of situations. It is more than a stop-gap solution. It’s a real improvement.”
Languages
He says the service will charge a “minimal” monthly subscription to cover the cost of the high-quality app and the outbound minutes.
The service, which will be available in a variety of languages, is likely to be strongly welcomed by users in Ireland, where there are very few options for text-to-speech services.
Only Eircom offers a text-to-speech relay service for deaf users, which relies on a human operator and requires a landline and a specialist device called a minicom, but is a "very underused and poor quality service", according to Niall Maguire of Deafhear, which has been lobbying Comreg to push operators to offer a better quality and more flexible text relay service for deaf users.
