Nuance tames IoT interface woes with new developer tools

15.12.2015
While the Internet of Things is gaining in popularity, makers of smart devices like robots, intelligent appliances and online-connected thermostats still need to straighten out their user interfaces. Right now, companies build physical interfaces into their products like dials and buttons, but those aren't great for executing complex commands like "go sweep the living room."

That's why Nuance is launching a beta version of Mix, a new service to help developers create devices that recognize voice commands and turn them into action. By using Mix, companies can focus on building hardware, and then leverage Nuance's years of speech-recognition expertise to let users control their products just by talking. 

There are two key components to Mix: the first is speech-to-text capabilities that convert end users' spoken words into text that can be fed into software. The second is a system that lets developers model how their users will interact with a device so Mix can pass commands from users' speech to an application.

The model developers create with Mix is specific to what they're building. They create a set of intents, and a group of concepts within those intents, so that a system can interpret "set my thermostat to cool for the next week" as a situation where users want to manipulate their thermostat and apply a particular setting.

Using an application-specific interaction model will help create better products. IoT devices don't need to recognize the entire range of English speech, just a subset of commands tied into what the device is supposed to do. 

Once developers have a model set up, it can then learn from users' speech patterns to improve and morph based on how people actually operate a product. Developers are in complete control over what anonymized speech data is fed back into their models, too. 

According to Kenn Harper, the senior director of mobile devices at Nuance, Mix will eventually be available in multiple pricing tiers, including one that allows developers to just pay as they go for what they need with minimal support, and another that includes a higher level of partnership between Nuance and an end customer. 

The company's bet is pretty clear: Nuance thinks that it can make a lot of money off companies that don't want to pay to develop their own speech capabilities in-house. It doesn't matter whether those are hobbyist businesses just starting to get off the ground and launching products on Kickstarter or if they're large, established players. 

What remains to be seen is how willing companies are to pay for the new capabilities (though Nuance Mix is free for testing purposes during the beta period) in order to empower their devices. 

Blair Hanley Frank