Forget the Mouse - Just Talk Instead

February 18, 1997

My computer is taking dictation today. I speak, it types out the words. At least it's trying to.

Before it got cleaned up, that second sentence said "it types OF the words."

But look beyond this initial mistake and something quite exciting is taking place. After years of using keyboards and the mouse, users are about to get a third method of controlling computers: their voice. In a few years, you may talk to your machine as routinely as you type today.

"We think we are on the front end of a whole new computer interface," says Barry Neagle, manager of North American sales and markets for IBM's speech software. (IBM's hot-selling VoiceType Simply Speaking program is taking the dictation for the first hamlet ... er, the first half ... of this column.)

"Speech will have as much of an impact on the way people use computers as the mouse and graphics did 10 years ago," says Roger Matus, director of marketing for Dragon Systems Inc. in Newton, Mass. "By the year 2000, people will be using speech as the primary way in which they enter information into a computer." (The company's DragonDictate software will take down the second half of this column.)

The year 2000 is awfully close. After using both these popular products, I think speech-recognition will become an important part of computing. It already is in certain specialized uses, such as medical-dictation programs and computer-telephone systems. But it will take longer than three years for the technology to become a fixture for the mainstream computer user.

For one thing, there's the microphone headset. To get the best results, both the IBM and Dragon Systems programs include a lightweight headset. Generally speaking, I wouldn't mind donning a headset to dictate several letters. But it's too much hassle for entering a few computer commands, like "Enter" and "Page Up." Another challenge: you have to wear the headset the same way every time. If the microphone is positioned a little differently, it can throw off the program's accuracy. Then there's the way you have to speak when dictating: in - sentences - with - small - pauses - between - each - word.

All the major speech-recognition companies are hard at work on consumer versions of the software that won't require those pauses. Maybe the headsets can eventually go too. It's easy to see how important speech-recognition could be for users of those small, hand-held computer devices, who currently make do with tiny keyboards and computer pens.

But are we at the point where you could dictate a letter to one of those hand-held devices in a crowded airport?

Know! ... er, No!

It's not fair to compare IBM's $79 Simply Speaking program with Dragon System's $695 Classic Edition of DragonDictate. They do different things. Strictly as a dictation system, the IBM product is slick and smart. Even with a small 22,000-word vocabulary (compared with the 120,000 words in DragonDictate), it performed quite well. DragonDictate's interface is not quite as straightforward, but it does more. Besides dictation, it allows users to open, close, and control by voice most of their applications, including word-processing, spreadsheet, and Internet browser programs. Again, the results were better than expected.

One of the impressive things about both programs is how well they learn. The first time through an excerpt from Dickens's "A Christmas Carol," they rendered Scrooge as "squeeze" or something similar. Corrected twice, they never made the mistake again.

Work with these programs and the dedicated user will find them useful. Wait for them to improve and the rest of us may one day be talking to our streets.

Er ... our screens.

* Send comments to lbelsie@ix.netcom.com or visit my In Cyberspace forum at http://www.csmonitor.com