|
If you flunked touch typing in high school, the
software industry may have the answer to your
typing woes: voice recognition software. Mind you,
voice recognition software is still not quite at the Star
Trek level of just barking orders at you computer and
expecting it to understand. First, you have to train
the software to your particular pronunciation. And if
you have an unusual speech pattern, such as a heavy
accent, or (as was the case while I was writing this
article) a nasty case of the flu, don't expect things to
go exactly as advertised. When you were a child, it
took you months to learn how to identify
particular sounds as words. If you're expecting your
computer to get everything correct after training it for a
half hour, well, you'll be disappointed. Still, the
technology has come a long way, and it is increasingly
useful.
I looked at two popular packages: Dragon NaturallySpeaking for use with WordPerfect
Office 2000 and IBM ViaVoice for use with Lotus
Word Pro. My selection criteria for these programs
was rigid—I had them already, as freebies bundled
in with the software suites. As it turns out, NaturallySpeaking enjoys a strong reputation in
the voice recognition market, and my experience
reaffirmed the validity of this high praise.
Setup and Training
At this point, no voice recognition software is
ready-to-go right out-of-the-box. Both of the systems
I looked at required a fair amount of set-up time. However, the process is relatively painless, as
both software packages walk you through training
using well designed wizards. For each package, you
must, at a bare minimum, adjust the microphone. (Here's
a tip: if you're a mouth-breather—remember I had
the flu—be sure to position your microphone to the
side of your mouth so it doesn't try to interpret
your breathing as words.) You also need to adjust the
volume, and train the program for your specific
voice. Training the program means you must read
specific text aloud. The program then records and
analyzes that information. To improve accuracy further,
both programs let you expand your vocabulary by
importing documents and correction information.
As far as ease and speed of setup, Dragon NaturallySpeaking wins in terms of being
slightly friendlier to use and more engaging (I got to read
a Scott Adams' Dilbert essay to train the
program). ViaVoice does allow you to get away with less
reading, but then takes its own sweet time analyzing
that data. For either software package, you would be
well advised to allow a minimum of a half hour,
and, more realistically, a full hour to install, train, and
familiarize yourself with the basic commands.
Dictation
Once the software is up and running, then comes
the fun part: dictating. The basics are easy; you turn
the microphone on and start speaking. Note that
you must specify any necessary punctuation. So to get
I think, therefore I am. to appear, you would have to
say "I think comma therefore I am
period."
If you have a cold or an interesting accent,
you can look forward to doing a fair amount of
correcting. I twiddled around with both programs,
and couldn't get an accuracy rate over 80 percent.
However, my cold put me at a severe disadvantage. CNET.com reviewed both programs and claimed
accuracy rates in excess of 90 percent. That
sounds great, until you realize that 90 percent means one
in every ten words needs to be fixed, which is a lot
of correcting. OK, I didn't say the technology is perfect.
Both programs only recognize specifically
worded voice commands, so you truly have to read the
directions to use the software. New on the market
is Lernout & Hauspie Voice Xpress Plus, which
reportedly allows natural language commands when it's
used with Microsoft Word. This means you may use
different phrasing to get the same results. Natural
language commands are likely to be included in new versions of all voice recognition software.
ViaVoice comes with a handy feature. If you
say, "What Can I Say?" it pops up a menu of
possible commands. The bad news is ViaVoice
command choices are limited. ViaVoice (in this bundled
version) gives you only the bare minimum of
possibilities. You can turn dictation on and off, turn
the microphone on, off, or asleep, or have the
computer read back what you've written. The reading
feature lets you select the type of voice (such as adult
female) and the tone (thoughtful, happy and so on). It's
an odd novelty that mostly serves to emphasize some
of the inconsistencies of the English language (for
example, natural is pronounced "na-TUR-al").
However, for the visually impaired, this feature could be
a useful add-on. Another problem with ViaVoice commands is that after you say the "Begin
Dictation" command to start dictating, your mouse and
keyboard don't work. So to correct an error, you have
to say "Stop Dictation" then highlight the problem
text using your mouse and editing options. There
aren't any voice commands to correct errors, which
leaves the visually impaired out of luck. If they make
errors, they may be able to hear the mistake on the
playback reading, but they won't be able to correct the errors.
NaturallySpeaking, on the other hand, leaves
the keyboard and mouse active while the microphone
is on. You can put the microphone to sleep if you
are interrupted (or, if like me, you talk to yourself,
and start wondering where the extra words came
from). You also can correct errors using voice-only
commands. You just select words and replace them.
Editing and Formatting
Editing and formatting is where
NaturallySpeaking outperforms ViaVoice. NaturallySpeaking lets
you use voice commands to move the cursor, select,
correct, edit and format text. ViaVoice only allows
limited, clunky corrections that are controlled in
part with the mouse, not with voice only. All other
text changes must be done using the mouse after
turning off the dictation.
NaturallySpeaking lets you use voice commands
to format text in a number of ways, including bold,
italics, underline, font, capitalization, alignment,
and bullets. The downside here is you have to get it
to recognize the command as a command. You
must pause in between dictating and issuing a
command. For example, to bold the words quick, brown
fox, you must first dictate the words, pause, then say,
"Select quick comma brown fox." Once you see that
the text is highlighted, you say, "Bold that." Once
you get the rhythm down, this process becomes
routine. However, some commands simply never worked
for me, including Set Format That. The program
never recognized the words as a command.
The Bottom Line
Overall, while playing with ViaVoice and
Naturally Speaking was entertaining, I'm more productive
with my keyboard and mouse than I am with my voice.
On a positive note, unlike most humans I've met,
finally there's someone who listens to me and does what
I want. Well, at least 80 percent of the time anyway.
Voice Recognition - Stand Alone Products
- Dragon Naturally Speaking Preferred 5
Dragon Systems Inc., A Learnout & Hauspie Company - www.dragonsys.com
- Via Voice Pro 8 IBM - www-4.ibm.com/software/speech/desktop/w8-win.html
Voice Recognition -Bundles
- Via Voice is part of Lotus Smart Suite
Millennium Edition 9.5 - www.lotus.com
- Dragon NaturallySpeaking is part of
Corel WordPerfect Office 2000 - www.corel.com
|
