Calling in the "Voice Squad"

By Jim Bray

The conversational HAL 9000 computer from "2001: a space odyssey" isn't here yet, but ever-improving generations of voice recognition software means it's getting closer all the time.

Voice recognition software lets you control your computer by speaking into it, for a variety of reasons. Depending on the package you use, you can do things like command it to open and close programs, dial your telephone, or dictate straight into your word processor.

More and more companies are offering such tidbits. Microsoft's new 900 MHz cordless phone includes voice recognition that lets you do basic functions like loading or closing programs and using the phone. I've been trying out a couple of voice recognition "engines" and have been quite impressed with what I've seen (or is it "said?").

Dragon Systems' "NaturallySpeaking" can be purchased as a standalone product or as part of the Corel WordPerfect 8 suite. I used the latter version, along with IMSI's VoiceDirect Continuous Gold. Both products say they'll let you blather on to your favourite Windows applications, including WordPerfect, MS Office, Lotus Notes and/or SmartSuite and Eudora E-mail.

Both products demand fairly hefty hardware and a good chunk of hard drive space, though. The Corel/Dragon thingy won't be happy with less than a Pentium 133 housing 48 Meg of RAM (64 for Windows NT) and a "typical" install swallows up 180 Meg of hard drive space. IMSI's product wants a minimum of a Pentium 150 with MMX, 32 Meg of RAM for Windows 9x (48 for NT) and 125 Meg of hard drive.

And don't forget a microphone! So don't install this fancy stuff unless your system's up to snuff…

After installing the software you have to train it, a rather tedious process in which you read a bunch of text into the computer. I preferred the DragonSystems' process because it let me read an excerpt from Arthur C. Clarke's "3001: the final odyssey," which seemed appropriate. Unfortunately, I was getting interested in the story when "Draco" was satisfied that it knew my voice. Now I have to get the novel to see how it ends…

Once the programs are trained, you have to train yourself: for word processing, for instance, you need to speak all the punctuation marks you'd normally type, as well as learning the commands for editing, error correction, etc. You're supposed to be able to dictate up to a hundred sixty words per minute, and perhaps you can once you and the software are intimate with each other's idiosyncrasies. I found I had to speak quite deliberately, though, which slowed me down so much that I could have typed the same stuff thrice in the time it took to be a "virtual raconteur."

And even then the software makes mistakes, some of which are hilarious. If you think spelling checkers produce some howlers, wait till you see your spoken words! Despite having reasonable diction (oops, the software heard that as "a reasonable addiction"), I still had to speak much more carefully than I would normally.

It's kind of like talking to your kids, except that the computer at least pretends to listen...

In all, this is a nifty way for non-typists, the handicapped, or professional dictators (that's "people who dictate" as opposed to Saddam Hussein or Jean Chretien) to operate a computer, but I don't think it's time to throw away my keyboard yet.

Given the evolution of these products, however, that day may be coming sooner than I had expected.

Tell us at TechnoFile what YOU think