In a blog post published Thursday, Amazon and speech startup Voiceitt announced a collaboration that aims to make Alexa more accessible to people with atypical speech. Voiceitt is an app that uses machine-learning and speech recognition technologies to help those with speech impairments communicate and be more easily understood.
In a press release, Voiceitt acknowledged the popularity of smart speakers and digital assistants like Amazon’s Alexa—despite their massive popularity, however, they can prove inaccessible to those with various types of speech delays. This can make the communicatively impaired feel as though they are excluded from using voice-first devices. Teaching a machine fluent speech is hard enough; teaching a machine to learn atypical speech is exponentially more challenging. Thus, Voiceitt “recognized the opportunity to expand its technology offering to facilitate not only in-person communication but also interaction with voice activated and controlled devices.”
“We’re excited to work with Amazon to bring the benefits of voice technologies to a broad segment of customers who, until now, may not have been able to enjoy these products,” said Voiceitt co-founder and CEO Danny Weissberg in the press release. “Voice technologies are increasingly mainstream, and this Alexa integration is testament to the growing awareness among major technology players of the importance of ensuring these technologies address the diverse needs and preferences of their customers, including people whose voices deviate from standard speech. Integration of Voiceitt’s speech recognition with a powerful service like Alexa further demonstrates Voiceitt’s value proposition in a rapidly expanding industry, and of our vision—to make speech recognition accessible to everyone.”
Voiceitt, headquartered in Tel Aviv, was born in 2012. The Alexa integration came about through mentorship via Amazon’s Alexa Fund initiative. Launched in 2015, the Fund’s mission is to, according to Amazon, “[invest] in companies that are innovating in AI, voice technology, frontier tech, robotics, and more.” Participants receive up to $200 million in venture-capital funding for product development. The inspiration for the Fund comes from Amazon’s belief that technological experiences driven by the human voice will “fundamentally improve” how people interact with technology.
In an interview with me, Weissberg explained the impetus behind Voiceitt is about empowering independence. The advent of voice-first computing has myriad benefits to people with disabilities, particularly in a smart home context. One of the chief advantages of a disabled person asking a virtual assistant for help is it lessens the reliance on others for help. These experiences go a long way in instilling greater senses of dignity and self-esteem—feelings elicited from heightened levels of autonomy.
The major downside is voice-driven interfaces aren’t usually kind to users whose speech patterns deviate from the norm. That’s what Voiceitt aims to solve.
“Before the Alexa-Voiceitt integration, many individuals with speech and motor disabilities whom we worked with were accustomed to seeking assistance, whether from family members or caregivers, for simple tasks such as turning on the TV or a light,” Weissberg said. “The ability to utilize an Alexa-enabled device independently for the first time using the Voiceitt app will significantly impact the day-to-day lives of these individuals.” He added Voiceitt represents a “life-changing milestone” for many.
How Voiceitt is trained isn’t dissimilar to how one trains a voice-driven UI like Siri, for example. It starts with a simple phrase, followed by repetition of words and phrases several times over. This helps Voiceitt learn the tone and cadence, among other qualities, of a person’s voice. The app then uses artificial intelligence to create a speech model with which it can process a user’s specific commands.
As for Amazon’s role, Weissberg said the company has been consistently supportive. Members of the Alexa team volunteered as mentors during the Amazon Accelerator program, and were excited by Voiceitt’s premise from its earliest days and committed to see the idea come to fruition. “[We] were inspired by Amazon’s commitment to ensuring that the benefits of cutting-edge voice technologies, which they were pioneering, would become universally accessible,” he said.
Peter Korn, who oversees accessibility at Amazon’s Lab126, is equally effusive of Voiceitt’s potential as an assistive technology. “We share [Voiceitt’s] vision to help people with speech impairments live more independently through voice,” he said in Voiceitt’s press release. “We were delighted to support them through an Alexa Fund investment and now through an Alexa integration via their mobile app.”
Weissberg told me early feedback from beta-testers was “overwhelmingly positive and quite moving.” He said testers have grown significantly more independent because of Voiceitt’s technology, accomplishing more on their own than ever before. A recurring bit of feedback was using the app has been a “life-changing experience,” he said. He added Voiceitt and Alexa has enabled people in ways they never imagined.
“Many [testers] report that for the first time, they’re able to join the millions of people around the world who are already using Alexa to control their smart home devices, listen to music and get information using their very own voices—something they never thought would happen,” Weissberg said.
One such tester is Nat Grupp, who’s been part of the Voiceitt pilot program since last year. Grupp, who’s in his thirties and lives in Philadelphia, has cerebral palsy. He has a passion for his hometown sports teams—the Phillies, Eagles, 76ers, and Flyers—and likes to keep tabs on them through the news and by watching games on TV. His condition has made it difficult to move and communicate; his father has long had to help him turn on the television. While Alexa initially was hard to use because the assistant had trouble understanding him, Voiceitt’s technology was a breakthrough. Gupp now can control his home’s lights, Drop-In on loved ones, and keep up with the Philly sports scene—all made possible by Voiceitt and Alexa’s software.
“It feels like a new superpower,” Grupp said in the blog post.
Voiceitt’s work in this realm is game-changing, insofar as it’s heartening to see people recognize speech delays are disabilities too and require accommodation. The mainstream tech press too often has myopic focus when it comes to assessing smart speakers and digital assistants: all they care about are their smarts. Barely any attention is paid to accessibility of any sort, particularly speech-related. Journalists and YouTubers pit Alexa, Google Assistant, and Siri against each other all the time to show how intelligent they are, which is valid in and of itself. At the same time, how well they literally understand you is taken for granted; these robots are built assuming normal speech patterns. The thing is, Alexa’s capabilities mean nothing if she cannot reliably parse your speech and do what you ask. Not everyone has the privilege of being able to effortlessly shout into the ether and have their thermostat adjusted. Some people have real problems with these tasks, as Weissberg mentioned. That he and Amazon recognized the need to wholeheartedly address these concerns is a mammoth development for the disability community—one that should be celebrated by the broader technology media. Intelligence is not everything for Alexa and her ilk.
For its part, Apple—as with silicon, the nigh-undisputed industry leader in accessibility—has tried to combat speech recognition issues with Siri in a couple ways. It released Type to Siri to iOS a few years ago, whereby a user can type commands and queries in an iMessage-style interface. While a good solution for certain populations, such as the Deaf and hard-of-hearing, the speech aspect of ambient computing is moot because typing doesn’t actually solve anything, it sidesteps it. As for tackling the actual problem of speech and digital assistants—it’s my understanding, according to sources, Apple has hired speech pathologists for its Siri team in the past to better understand the physiology of speech, and make Siri more graceful with atypical speech.
By contrast, Voiceitt and Amazon are actually taking a page from Apple’s playbook nowadays and applying a truckload of neural engine-y, machine-learning tricks to try to better the user experience in this regard. Maybe Voiceitt inspires Apple’s work.
Weissberg is excited for Voiceitt’s future, telling me “we never imagined just how much we would be able to achieve.” His team eagerly welcomes feedback from users and their families, as making speech-impaired individuals an “integral part” of voice-first computing is a priority. “Bringing accessible technology to people with speech disabilities, together with Alexa, is our main focus for the near future,” he said.
The Voiceitt app with Alexa integration will launch in Q1 2021 on iOS. It is available for pre-order on the Voiceitt website.