Voice Market for Children Is Ripe for Innovation
Children are exposed to technology at early ages. Parents often use devices to provide kids with educational and entertainment content. Meanwhile, voice recognition is emerging as a common way for adults to interact with technology and has the potential to create new options for children as well. Yet little work has been done to date with children’s voice recognition systems. “When my daughter was 4, I searched but found no children’s voice recognition products,” says Patricia Scanlon, founder and CEO of Soapbox Labs. Gradually, new solutions are emerging, but this market presents vendors with many vexing challenges.
The many differences between children and adults impact speech recognition system design. Youngsters’ attention spans are much shorter than those of adults. An adult will sit with a device and input information in a serial process, but toddlers often move from one item to another in a random sequence.
Also, adults understand that lessening background noise is a must; kids don’t. They bang on tables or touch microphones, which impacts their input.
In addition, children approach technology with different levels of development when it comes to spelling, writing, and keyboard usage. Voice recognition has to fill in the blanks and compensate for information minors do not input.
Data Privacy Red Flags
Voice recognition requires companies to collect large samples of conversations, develop algorithms to recognize information in context, run tests to determine effectiveness, and then tune the system. But collecting juveniles’ data has raised red flags. At the Consumer Electronics Show (CES) 2017, Mattel debuted Aristotle, a voice assistant aimed at children from infancy to adolescence. Resistance came quickly from the Campaign for a Commercial-Free Childhood and Story of Stuff Project, and the system was shelved.
Complicating things, the laws governing data collection are unclear, constantly evolving, and usually behind the technology. Last fall, the U.S. Federal Trade Commission mandated that recordings without parental permission must be destroyed.
In Need of More Skills
Consequently, the children’s voice ecosystem has been built up slowly. In August 2017, Amazon launched Alexa Kid Skills, which collects parental permissions via the Alexa app for other voice-enabled solutions. Amazon Storytime, Animal Sounds Quiz, Old Macdonald, Sesame Street, SpongeBob Challenge, and Zoo Walk are some of the applications that have been developed using this foundation.
A few start-ups moved in to make it easier for third parties to build applications that rely on children’s speech recognition. Soapbox Labs landed $1.2 million in seed investment from Astia, Elkstone, and Enterprise Ireland as well as private investors. The company developed a speech recognition system, built on a deep neural network, that relies on a data set of 50,000 children’s voices in 150 countries to determine a youngster’s input.
The vendor wrote a cloud-based API that includes sounds, words, sentences, fluency, and comprehension features. Soapbox licenses its technology to third-party developers that voice-enable reading applications, robotics, language learning solutions, and smart toys.
Tell Me a Story, Daddy
Novel Effect, a Techstars-backed start-up, developed a mobile application that works on Apple iPhones and provides interactive book reading capabilities. “The application is designed to have the parents and children read together,” says Matt Hammersley, Novel Effect’s CEO and cofounder.
The app’s voice recognition system does not rely on sampling. “Our voice recognition system has a different design than traditional systems and works with any type of voice input,” says Hammersley, who declined to provide more information about the system.
To date, vendors have focused on improving adult speech recognition, but with the profound differences between adults and children and the regulatory hurdles involved with collecting minors’ speech, a void existed with children’s speech recognition. Solutions are starting to emerge, but much work remains before they are widely adopted.