Wednesday, November 19, 2014

Like the hoverboard, Speech recognition is one of those technologies that hasn’t quite lived up to its promise. We’re all capable of imagining a future where speech recognition software is sophisticated enough to recognize the entire spectrum of our speech patterns and respond accordingly, it’s just not here right now.

 

In fact, it’s still a very long way off, according to speech recognition expert James Kendrick. In a recent ZDNet report, Kendrick says we’re no closer to accurately interpreting the human voice with a computer than we were a decade ago. Apple’s Siri is not accurate enough to make anything longer than a brief message worth uttering - and even then it often takes so many repeated attempts it would have been quicker to just type.

 

According to Kendrick, “even speech input in a totally quiet environment using a high-quality noise-cancelling headset only gets you 90-95% recognition accuracy.” Not good enough to start dictating 800 word essays without a laborious editing process.

 

There have been a few notable improvements to IVR technology that at least suggest a resurgence of interest in developing truly precise speech recognition. A good IVR solution is invaluable to businesses with high call volumes. They have a strong interest in supporting research and development into the improvement of voice recognition tools.

 

Kendrick concedes that much of the hardware has gotten significantly better. It may just be waiting for the right flash of genius from software developers to really take us to Star Trek territory.


As flippant as that may sound, popular sci-fi goes a long way towards keeping outlandish technological dreams alive, and often shapes the way our innovations look and function. They keep the flame of our imaginations fueled during R&D lean spells. The tech industry and consumers alike are looking forward to the day when 100% accurate speech reconition goes from fantasy to reality...