It’s easy to joke about digital assistants like Alexa not understanding us. But that’s because most of us take for granted that the words we speak are easily understood. For many people with non-standard speech patterns, this isn’t the case. People with cerebral palsy, stroke victims or people with lifelong speech impairments may speak in a manner that digital assistants, or even other humans can’t readily understand.
Tel Aviv, Israel startup Voiceitt is working to help fix that with its mobile app that works as an individualized electronic translator. Once installed on a phone or tablet the user trains the app by speaking different words. Once Voiceitt has learned those words, the app can then be used to “speak” on behalf of the user. The technology is voice agnostic, so it can work in whatever language the app is delivered in.
You can see it in action in this video, where the user has trained the Voiceitt app to understand how he pronounces “burger” and how that means he would like a burger served in a particular way (with cheese, etc.). When it comes time to order, he says “burger” and the phone’s loudspeaker translates the order in a way that is easily understandable by the server.
Voiceitt also ties in directly with Amazon’s voice assistant, Alexa. (Voiceitt is funded in part by the Alexa Fund.) In this way, Voiceitt skips the intermediary step of saying the translation out loud on the phone. Instead when a command is spoken (“Turn on the light”) Voiceitt’s app talks directly to Alexa through the API to execute the command. Turning on a light may not sound like much, but without voice control, people with cerebral palsy, for instance, are reliant on caregivers to come in and do basic tasks like turning on a light or adjusting the volume on the TV or radio.
For our purposes here at The Spoon, this also means that a person with non-standard speech could also operate connected appliances like microwaves or water faucets, or order food and groceries for home delivery.
As noted, right now Voiceitt requires each user to train the app individually. It can’t extrapolate to words beyond the trained vocabulary spoken in that particular manner. But as more people use the app, Voiceitt’s database of speech patterns grows. Over time, Voiceitt’s artificial intelligence will process its library of data, recognize more patterns to become more fluent and more of a universal translator.
As that happens, Voiceitt could then be installed on the business side of more locations. For instance at a restaurant’s drive-thru, or voice-controlled ordering kiosk. With the software built into the kiosk, users wouldn’t need the middle-man of the tablet or phone, the kiosk would translate.
Voiceitt has raised $16 million in funding so far including non-equity money from governments and other funds interested in bringing more equity to technology. The Voiceitt app will be available here in the U.S. at the end of May, with a subscription of $200 a year, but Voiceitt says it is working to partner with relevant agencies to make it more accessible to people who aren’t able to afford that cost.