Frankly speaking

Leftover thoughts about things for which to be thankful after the really important ones have been named: GPS – the fact that it knows where I am and can guide me to my destination and back again, pronouncing almost every street name and landmark correctly. We all but take this technology for granted, but it’s a big improvement over directions scribbled on the back of an envelope or, “two lefts, then a right, then another left…”

I still enter my destination the old-fashioned way – on a touch screen – but chances are my next upgrade will respond to voice commands. An announcement from Nuance Communications that personal navigation device (PND) maker Mio Technology would make greater use of the Nuance Vocalizer for Automotive system brought voice to mind.

Arnd Weil, Nuance’s general manager, Automotive, says the firm’s customer list includes most major PND developers and more than 25 automotive brands. It’s no longer sufficient for Nuance to provide speech technology, Weil says, although the firm is working to improve accuracy depending on what a speaker says or how he/she says it.

Preparing Recommendations

With relatively limited hardware, Nuance needs to interpret commands spoken in myriad regional and national dialects and not to be stumped if a user speaking, say, German, requests a song in English (Ford’s SYNC uses Nuance technology).

Weil notes that connected services are becoming increasingly important, and that trend has implications for Nuance and other speech recognition firms. “Many, if not most devices will be connected,” he says. “There will be more information available, and drivers need to be able to control that information in user-friendly ways.”

Connected applications, whether for PNDs or factory installed systems, are limited by developers’ imagination and the combination of functionality and user interface. Weil says Nuance is increasingly involved in application and human-machine interface development. It’s leveraging work done by other divisions to benefit automotive applications; for example, a text-to-speech screen reader originally developed to aid blind people.

But developing speech recognition applications has to be a huge challenge. Presumably the best applications are those able to recognize and respond appropriately to the words that drivers are most likely to speak in a given situation; for example, “(Gee whiz), I’m late and I’m stuck in traffic. WhaddamIgonnadonow?”