The Soundboard


Editor: Jean Van Loon
Heritage Village Computer Club
Web Edition - April 2000

Website: www.heritagevillage.org
E-mail: theweb@freewwweb.com
Color
Speech Recognition Software

Patience and Computer Power Indispensable

So you would like to start talking to your computer. First you will have to figure out what program to get, and there are many to choose from. People use speech recognition for a variety of reasons. They may wish to be more productive or to accommodate disabilities. Maybe they cannot type.
In any case, speech recognition requires an investment of both time and money. If your computer is more than a couple of years old, you will probably need more up-to-date hardware to run a speech recognition software program effectively. It takes time to decide on a program and then more time to learn how to use it.
Dragon Systems, IBM, Lernout & Hauspie and Philips have a number of products on the market ranging in price from $60 to $600. These companies also offer special applications for lawyers and doctors as well as packages that include handheld recorders. All programs take dictation in continuous speech which, despite company hype, is not quite natural. Today's programs do write down what you say, as fast as you can speak clearly, with a high degree of accuracy. But these levels of proficiency take practice by training on the program that you select.
The training is essential. You cannot expect to go to the computer, start talking to it and have your words accurately transcribed into text. After initial training, which takes from 5 to 10 minutes, you will learn to

improve the program's accuracy as you dictate, by correcting its mistakes in transcription.
Selecting a program is a somewhat daunting task as each one has different options beyond deciding how much you want to invest.

Guide for Using Speech Recognition

When you do the initial training, speak easily and naturally. If you use a best behavior voice the software will expect it from you all the time.
Pronounce each word distinctly, moving your mouth and lips naturally.
Don't just delete mistakes made by the recognition engine; use the specified correction process for teaching the speech engine.
Position the microphone properly and consistently. It should be placed at the side of the mouth and slightly below it - a thumb's width to two fingers away. This will eliminate the unwanted recognition of breath sounds.
Some programs have a control to balance speed versus accuracy. Push it toward "Accuracy" and see if the loss of speed is acceptable.
To eliminate commands being transcribed as text, pause between dictation and entering a command. Don't pause in the middle of a command or it will be entered as text.
Develop a fluid speaking style. Slow, halting speech is not well understood. Long phrases are better under-stood than short phrases and short phrases better than words. Don't wait to study your results, just keep dictating.

KEEPING PACE IN CYBERSPACE
Select Page to View
1 | 2 | 3 | 4 | 5 | 6 | Home