|
by
Matthew Stoton
When IBM ViaVoice
first appeared on the scene, the first generation of
speech recognition, it was hailed supreme. Not surprisingly,
since it was the only continuous speech recognition
software available for the Mac. Now the second wave
of speech software has landed with IBM ViaVoice
2.0 and iListen
1.0. This next generation is aimed at dictation
and control. By allowing you to control and dictate,
this software helps you keep your hands off the mouse.
iListen's speech engine (licensed from Philips) is advertised
as faster than IBM's, but can it outshine the revolutionary
ViaVoice?
You Listen: Installing iListen is a breeze,
though slightly dull. But beware: it won't work on OS
X and stability on 9.1 varies, it is recommended that
you use 9.0.4. When you first open iListen you are asked
to make a profile and setup your microphone and voice
model. After naming your profile you must set the gain
and "Silence Detection Level", or SDL. Setting
the gain can be tricky; set it too high and signal quality
depreciates, too low and it can't hear you. You can
have it set this automatically by listening to you but
it may fail if you talk unevenly. For most people it
should be set to about 50% to 75%. It ignores sounds
below a certain level so that things like breathing
and background noises are ignored (the SDL). The SDL
should be about half of the gain. So if your gain is
66% SDL should be set to 33%. This may vary depending
on your voice and environment.
One more story, Please: The most annoying part
of today's speech recognition is reading stories. This
not only goes against the idea of "you talk, it
types", but is also time consuming and relatively
boring. iListen brings shorter and somewhat more interesting
stories. Even with the promise of 85% to 95% accuracy
after the first story, it is wise to read them all.
Without the extra information the software will blunder
on most words, making dictation nearly impossible. Still,
command mode works beautifully regardless of how many
stories you have read, and spell mode is also unaffected.
Don't worry though, you will be saved from reading pages
of "Treasure Island". iListen gives you a
mixture of short and long stories, most of which relate
to speech recognition. Reading all the stories is the
only way for dictation to shine; otherwise performance
may be less-than expected.
Commanding or Controlling?: Apple's feeble
attempt at command-control software isn't exactly beautiful.
Even though iListen is dictation software it sports
a powerful command-control mode that almost outshines
the dictation. It uses the Philips' FreeSpeech 2000
engine instead of Apple's PlainTalk. This allows it
to overcome the common difficulties associated with
PlainTalk technology, such as getting bogged down after
several commands and generally poor performance. With
iListen, commands are recognized instantly and missed
commands are nonexistent (if settings are correct).
Menu control isn't working yet and therefore not in
iListen, it should be working in version 1.1 or 1.2.
Spell Me: Ahh, Spell Mode, that delightfully
clunky version of dictation. In IBM ViaVoice Spelling
Mode is used for uncommon words or for correction. In
iListen this is used mainly for those words that iListen
doesn't know. Since there is no way to teach iListen
words, Spell Mode becomes somewhat overused. As usual,
it doesn't work incredibly well and is no better than
IBM's. Hopefully, when learning commands becomes possible
(in version 1.1 or 1.2) it will make dictation less
reliant on this clumsy feature.
Toss that keyboard!: Using voice to replace
the keyboard is the goal of dictation software, and
iListen almost does just that. iListen can dictate faster
and more accurately than IBM ViaVoice Millennium Edition,
thanks to the FreeSpeech 2000 engine. iListen's dictation
can work inside any application, but this comes with
a cost. iListen has no correction features, forcing
you to reach for the keyboard every time you make a
mistake. As long as you train it enough, it won't make
many mistakes keeping keyboard usage to a minimum. This
problem will be fixed in version 1.1 or 1.2 which will
be a free update, so it is no reason not to buy this
version. In short, dictation is much faster and more
accurate than what IBM has to offer.
To Talk or to Type: In the end, iListen is
better than IBM ViaVoice. Even when considering the
youthfulness of this program it comes on top in almost
every way. A bigger decision is to talk instead of type,
and it does take time for both you and the software
to adjust. iListen will inevitably increase your productivity
and ease your wrists. It isn't perfect, lacking OS X
support for example, but it is definitely worth it.
iListen finally proves that speech recognition is not
just for those with wrist injuries. This is the first
step towards keyboard and mouse elimination, lets just
hope that it stays on course.
Product: iListen
1.0
Company: MacSpeech
MSRP: $139......5/21/01
Hits: Fast, accurate, reliable dictation software.
Misses: Dull interface, no correction mode, lack
of Mac OS X support, flawed spell mode.
Rating: 4 out of 5
Matthew Stoton is a freelance
writer, webdesigner and software developer.
You can view his work and more at his site, Totally
Macintosh. He also offers
free technical support. You can reach him at:
rjsanddls@aol.com
|