Voice Dictation Software

The other type of voice-recognition software is far more complex. Converting standard speech into
text is an extraordinarily difficult task, given the wide variation in human speech patterns. For this
reason, nearly all software of this type (and some of the basic voice command applications, as well)
must be “trained” to understand a particular user’s voice. You do this training by reading prepared
text samples supplied with the software to the computer. Because the software knows what you’re
supposed to be saying beforehand, it can associate certain words with the manner in which you speak
them.
Users’ results with this sort of application vary widely, probably due in no small part to their individual speech patterns. I’ve heard people rave about being able to dictate pages of text without touching
the keyboard, whereas others claim that correcting the many typographical errors is more trouble
than typing the text manually.
Many variables can account for such variances in the quality of voice dictation. Early versions of these
products required users to separate words and speak “robotically - like - this” to get the computer to
recognize the text. These “discrete speech” products are now obsolete and should be replaced by
newer “continuous speech” products.
A second factor is the “trainability” of the software. My experience with trainable and “no-trainingneeded” programs suggest that training a voice-dictation program to recognize your speech patterns
and accent is still a good idea for best results.
A third factor is the program’s active and total vocabulary. Given identical voice-recognition engines,
the program with the larger active vocabulary will react more quickly to dictation, and the program
with the larger total vocabulary will be capable of storing more user-specific words.
If you’re involved in medicine, law, or other professions with specialized vocabularies, you might find
it saves time (and ultimately money) to purchase industry-specific versions of your favorite voicerecognition program. These programs offer enlarged vocabularies tailored to your line of work.
Other features to look for include voice commands for computer operations, text-to-speech translation to let your computer “read” to you, and a Web browser. Mobile versions of these programs use a
special digital recorder to save your dictation when you’re away from the computer and allow you to
play it back for voice recognition when you return to the computer.

Whatever the program you choose, some prerequisites for success include the following:
■A high-quality sound card (compatible with the Sound Blaster 16 or better); check the software
company’s Web site for the latest list of recommended sound cards.

■A quiet room with little change in background sound.
■A microphone specially designed for voice recognition (a low-cost one normally comes with the
software, but better models are available). Using a poor-quality microphone or an inadequate sound
card will cause problems for even the finest voice-dictation program. For best results, use a microphone with noise canceling or directional features, and choose a model that’s on the approved
list compiled by the voice-recognition software vendor. These usually feature a headset for precise adjustment of the microphone-to-mouth distance and angle, which is a crucial element in
good voice recognition.
■Completing sound setup and user training, if offered, to set software for your voice and vocabulary.
■Retraining difficult-to-distinguish words.

■Using the most advanced version of the software you can afford.

Voice-recognition technology is still a young technology and is sensitive to changes in a person’s
voice. Illness and stress can often change a person’s voice enough to throw off most of the “consumer” voice-recognition products. You will also need a fast computer, at least a Pentium 200MHz
with 64MB of RAM or more, to achieve response times that are quick enough to keep up with your
speech. Because voice-recognition technology (especially voice-dictation) can also help users avoid the
debilitating effects of repetitive stress injuries on wrists, fingers, and arms from typing too long without a break, it should become increasingly popular as system speeds increase and the software

improves.

Note

As with any resource-hungry product, your system should exceed the “minimum requirements” for RAM, CPU speed, and
free disk space by a factor of two or three if you want to be really happy with your software.

However, advances in this technology are rapidly developing to a point where more and more of us

are able to use continuous speech to write with computers, rather than typing.

Voice Dictation Software