Killer App for voice recognition – don’t hold your breath

1999-09-08

IBM’s newest version of ViaVoice signals a change of tactics for their voice recognition software. The product is to have improved recognition, better user-friendliness and additional features including Web surfing facilities. Will this release prompt the acceptance of VR? Unlikely, not yet.

There are three areas which we can see as benefiting from this technology. The first is dictation. Speaking, ultimately, is a different communication technique to writing. Some dictate letters, but most prefer to write them directly. Will this change? Unlikely, particularly in today’s already noisy offices. Imagine the hubbub of everyone talking inanely to their machines. As I write this, I contemplate using a microphone rather than a keyboard: even with 100% recognition, I cannot see the attraction. Maybe that’s just me. The second area is to provide an interface for users with physical disabilities. There is a clear benefit here, to certain users. Even so, neither of the two areas are sufficient to give killer app status to recognition software.

The third area is in commanding the computer. This is where the potential lies, but unfortunately not in its present state. The ability to launch an application, or even surf the Web for that matter, by voice commands is unlikely to be any more than a fad. Voice recognition will sit on the bench, as it waits for the technology which will launch it into ubiquitousness: intelligent agents. Once computers are capable of following natural language commands, then the time will be right for their communication by voice. For computers, read the whole of the online infrastructure. For example, “Take me to Amazon. I want to buy the latest John Grisham novel,” is an sentence I could imagine speaking out loud, either to my computer or directly to my phone.

The issue is one of time. The recognition software quality is almost there but the ability of the IT infrastructure to make sense of our utterances is not, in terms of both indexing the information “out there” in a sensible fashion and providing the resources to exploit it. Neither is it a priority, so it is likely that voice recognition has a wait on its hands before its perfect partner comes along.

(First published 8 September 1999)