The Internet of Things: A New Evolution for Speech Technology

Bouzid, on the other hand, asserts that the more focus put into proper UI design, the less machine learning has to be employed. “The challenge is very basic right now,” he says. “It is important for voice designers to define and follow best practices for UI. This means making long prompts shorter, clarifying language, and reducing confusion.

“It’s ultimately twenty percent of the causes creating eighty percent of the problems,” he continues, alluding to the Pareto principal, “and designers can produce much higher-quality experiences quickly by addressing these issues, whereas machine learning has to reach a critical mass of data and has to be trained in order to reach its full potential.”

Groups like the Web of Things Interest Group at WC3 are looking to explore voice’s potential, define and implement standards for the IoT, and define best practices for speech UIs. Following suit, VoiceBox Technologies has partnered with Macquarie University in Australia to employ AI techniques in advanced semantic language understanding, large-scale data mining, deep neural networks, and advanced machine learning, and it plans to expand such research and development partnerships to other regions.

Developers looking to include speech outside of the proprietary environment may find tool kits such as wit.ai, Nuance Mix, Microsoft Louis, and IBM Watson Developer Cloud attractive for prototyping or for building the appropriate architecture for a suite of speech commands.

The Internet of Things Market

An early locus of the Internet of Things was the car, and the automotive industry’s contributions to the IoT remain strong. Speech is the prevailing UI in the automotive industry as well, making it a prime space for speech providers looking for partnerships with companies wary of proprietary tech martinets like Apple, Microsoft, or Google. Auto manufacturers Hyundai, Honda, Ford, Audi, and Cadillac all offer voice-control options and internet-dependent features such as navigation. With few of these companies eager to align themselves with Apple, Google, Amazon, or Microsoft, the market to provide speech technology for them remains active and a source of opportunity for small and large speech vendors and developers alike.

While home control is where speech technology most shines in the IoT, other IoT frontiers seem less likely. The IoT in B2B and industrial use cases is almost universally under the hood, running in the background to monitor equipment, automatically order supplies, or alert personnel of malfunctions. Offices, public spaces, museums, airports, and classrooms have yet to produce promising use cases for speech due to issues with background noise, simultaneous engagement from multiple users, and a desire for centralized control of typical IoT items such as thermostats and lighting. Dahl, however, suggests that an intriguing use case for speech in public spaces may be found in hospitals.

“Speech as an accessibility tool is easily overlooked,” she says. “The IoT could be used to provide quality speech recognition and control in spaces where patients have limited mobility for functions like dimming the lights, adjusting the bed, controlling entertainment, or calling for assistance”—all applications that would be exceptionally attractive to a medical industry with overworked diagnostic and nursing staff.

Remaining Problems

While speech stands to benefit from a new age of the Internet of Things, its new environment poses additional challenges and exacerbates old ones.

“The Internet of Things is making transitions between interfaces worse,” says Melfi, who emphasizes that the IoT will likely be a multi-interface landscape of touch screen and speech. “There is currently no provision for devices like the Echo or your smartphone to know how many people are in the room,” Melfi states. “Currently, there’s a one-to-many relationship with the IoT—one user to many devices.” Products like the Amazon Echo, Apple Home, and Google Home may be trying to reduce the equation to one-to-one by offering a single control hub for a multiplicity of devices. But Melfi insists that the future of the IoT will be many-to-many.