Voice recognition software now a viable option

by Alan Zisman (c) 2002 First published in Business in Vancouver ,  Issue #  date  column

Many computer users haven't realized that they can now tell their computers to "take a letter."

Recent versions of Microsoft Office have included the option to let MS Word take voice dictation. This option is not installed by default, and few have used it. I haven't, but all the reports I've heard suggest if you really need to be able to talk to your computer, you're better off purchasing a dedicated voice-dictation application, such as Dragon Systems Naturally Speaking or IBM ViaVoice.

I spent some time with the latest edition of ViaVoice Pro USB Edition.

This version includes a high quality Plantronics headset connecting to your computer via a USB port. This provides better audio fidelity than typical headsets connecting to the sound card, and better fidelity equals more accurate voice recognition. If you want your computer to understand you, make sure you use a good quality microphone. Voice dictation performance also benefits from more powerful computers. The Mac version makes good use of Apple's more powerful G4 processor-line, though it will work with G3 models.

Start off by training the software to work with your voice and accent. This takes about 30 minutes reading from a series of sample stories. You can add additional users, with each new voice taking up a whack of hard drive space. (You may also want to create separate profiles for different environments: a quiet home setting vs. a noisy office, for example.) The software's accuracy can improve. When using the built-in SpeakPad application, it learns from your corrections.

Once enrolled, you can use ViaVoice in several ways. Along with using Speakpad, you can dictate to any application that accepts text: word processors, e-mail, and more. This works well though ViaVoice had trouble with phrases that I slurred together. All it takes is a pause to switch between dictating and speaking one of the 500-plus pre-programmed computer commands. The program is smart enough to know that the words "File Save" are meant to be a command and not dictation.

The command "Surf the Web" opens your browser. Commands such as "Scroll down" or "Back" and dictated Web addresses work with Internet Explorer or Netscape Navigator.

As well as taking dictation, ViaVoice can read your text back to you in a mechanical but understandable voice. (The Mac version can make use of the wider range of voices included with both new and older versions of the Mac OS, sounding less robotic than the Windows version.)

In addition to a 160,000-word basic vocabulary, several specialized custom vocabularies are included in the package: computers, business and finance, cuisine and Internet chatter's jargon. The computer's vocabulary automatically translates "gigabyte" into "GB," for example. Specialized add-in medical and legal vocabularies are also available ($232 each).

Voice dictation has come a long way since the early days when users had to separate each word with a slight pause. It can now be a useable and attractive option for users unable or unwilling to type, though I doubt it will become popular in crowded offices.

If your voice-dictation system needs are more complex than those that can be met by simply installing an out-of-the-box program like ViaVoice, check with Vancouver's Speakeasy Solutions (www.speakeasysolutions.com). It offers consultation, installation, and customization of speech-recognition systems.

ViaVoice is available in versions for Windows 98SE or later, and the Mac (separate classic OS and OS X versions). It's also available in a number of versions. I looked at the ViaVoice Pro USB Edition ($280 Mac OS X, $310 Windows; other versions start at $47). There's even a new version for PocketPC hand-helds.