Full Disclosure: Voice Recognition Is Ready for Prime Time
It may be a little premature, but I already have a pretty good idea about my nominee for next year's World Class Product of the Year. I'm not entirely sure which product I'll root for, but I'm almost positive it will have something to do with broad-vocabulary speech recognition.
Yes, I know the old joke: Useful speech recognition is just three years away--and has been for two decades now. Even these days, I suspect, the technology will be slow to catch fire, mainly because people won't believe that it can work as well as it does. But two products have shown me that speech recognition (and I don't mean just the limited-vocabulary stuff that phone-mail-hell systems use) has become downright powerful.
For years, Dragon NaturallySpeaking (now owned by Scansoft) has been the best voice dictation product. Now version 8 of the application fulfills its predecessors' promise by delivering lots of little improvements that add up to much greater usability. Not only does the software present a more-useful-than-ever list of alternatives when you tell it that it has made a mistake, but now it allows you to correct your own verbal mistakes by selecting the error and then saying what you actually meant.
A new auto-punctuation mode attempts to insert periods at the ends of sentences, but it's so limited and inconsistent that you're better off sticking with the old-fashioned practice of speaking all the punctuation marks. And it's a shame the software can't automatically return you to the spot where you were working before you stopped to correct an earlier word. On the other hand, the software lets me talk fast, even on a five-year-old machine. And more than once I've dictated a quick draft on my notebook; the software's included noise-cancelling headset now goes into my traveling bag on every trip.
NaturallySpeaking Professional has a distinctly corporate focus and costs $700. But for most people, the $190 Standard or $180 Preferred edition will work just fine.
Right as I was beginning to realize how cool Dragon is, I spent some time with VoiceSignal's new VoiceMode software. Unlike NaturallySpeaking--which lets you talk pretty much the way you normally do, except for the punctuation--VoiceMode requires that...you...say...every...word...separately. Amazingly, though, it works not on a PC with lots of RAM and a fast CPU, but on a lowly cell phone that costs just $80 with a two-year contract.
Thanks to its built-in VoiceMode, Samsung's P207 phone, which is available from Cingular, allows you to speak text messages--and it's sufficiently accurate to be genuinely practical. Dictation would be even more welcome in an advanced handheld that can do e-mail, but even in this relatively basic phone it can free your thumbs from agony. In addition, the P207 lets you dial a contact's number from your address book by saying that person's name--a less impressive trick, perhaps, but a valuable one nonetheless.
The standard boffo ending to a column like this is the revelation that it was dictated, not typed--and indeed, I wrote this with NaturallySpeaking. But we're getting to the point where voice recognition, in an array of products, should be considered just another way to work faster and better. World Class, I'm talking to you--and I don't mean in 2008.