Issue #425 Business in Vancouver

The High Tech Office: Computers that listen when you talk... Dec 16 1997
by Alan Zisman

Despite all the changes in office technology, most of the time, most of the data that most of us input into our computers comes from the tips of our fingers, on our keyboards. Mice, glidepads, trackballs and the like are great for pointing and clicking, but simply aren?t an option for getting that report into print.

And that keyboard-dependence is one of the big obstacles in the way of getting more people to use computers. By now, most of the people who are comfortable typing on the job already have a computer on their desk. But if you can?t or won?t type? Maybe you aren?t physically able to work with a keyboard. Or maybe your job requires you to have things in your hand (radiologists, for example, scanning x-rays). Or maybe you entered the workforce at a time when the expectation was that management shouldn?t learn to type?that?s what clerical staff are for.

Well, there?s typing tutor software, to get users over the home-row hump. Effective products such as ?All the Right Type?, though that software?s Burnaby distributor, VR-Didatech aims it primarily at the education market.

Or pen input?a few years ago, heralded as ?the next big thing?, but now, despite big improvements over early efforts that made it the butt of a week of Doonesbury episodes, it?s settled into a small niche.

Many of us have been waiting for effective voice recognition, as a way to lose those calluses on our fingertips. There?s been voice recognition software for some time?IBM, for example, has been working in the field for over two decades. First attempts,  however, required expensive dedicated systems. And like the problems trying to make sense of our many different handwriting styles, our manners of speaking are often problematic for any hardworking computer.

Until recently, voice recognition software required discrete speech?speaking-with-a-slight-pause-between-each-word-so-that-the-computer-had-a-better-chance-of-understanding. A few years ago, I looked at IBM?s Voice-Type Dictation? it demanded discrete speech, and came with an add-in card, to beef up the computers of the day. It worked as advertised, but was too expensive and too awkward to use to gain widespread acceptance.

Since then, affordable computer power has made speech recognition more practical?and more affordable. The latest generation of products, such as IBM?s ViaVoice ($159), breaks through the earlier products limitations. No add-in card required, though you do need relatively hefty, modern hardware: an MMX Pentium 150 or better, with 32 megs of ram, and about 100 megs of free drive space. A sound card, and a CD-ROM drive. A headset microphone is included in the box.

And no more artificial pauses. ViaVoice promises support for continuous speech? speak in a moderate, conversational tone.

For best results, some training is needed. There are several levels of what IBM refers to as ?enrollment??the more time you spend letting the software get used to your mannerisms of speaking, the more accurate it will be. A minimal level takes just a few minutes, but for better results, be prepared to spend up to an hour or so reading as many as 265 sentences?including a ghost story written by Mark Twain. Then leave your computer alone for about an hour, to think things over.

Afterwards, it?s ready to take a letter. A simple word processor, IBM SpeakPad, is included, and if desired, dictation support can be added to Microsoft Word (versions 6, 7, or 97). Click on the Begin Dictation toolbar item, and start speaking into the microphone. If the program makes mistakes, double-clicking on the word brings up a correction menu, and lets ViaVoice learn better how to work with your voice.

Multiple users are supported?each will need to go through the enrollment process. The process isn?t perfect, however. Take a look at the opening sentence in this column again. When I dictated it, here?s what appeared on the screen: ?spinal the changes and computing most of the time most of the David that most of as input into arms she comes from working in tips on the people.?

Oh well? I only just installed it this morning. I guess I?ve got to speak a bit more clearly, and keep on training with the software.

In fact, when I tried that first sentence again, it came out: ?Despite all the changes in office technology, most of the time, most of the data that most of this input into or computers, the tip of our fingers,on are keyboards .? Still not perfect, but better. I suspect the software is training me as much as I?m training the software. The concept of being able to dictate to our computers is so promising, and ViaVoice seems so close to making it happen, that it seems worth the effort to persevere.

Search WWW Search

Alan Zisman is a Vancouver educator, writer, and computer specialist. He can be reached at E-mail Alan