IBM ViaVoice

IBM has been quite active in the area of speech recognition,
both at the desktop and enterprise level. Its ViaVoice speech recognition
software is available in many versions–Pro, Standard, Web, etc. There’s
another edition for the Mac as well. We checked out a special edition of the
product that comes with the Sound Blaster Live! Platinum card.

Installing
the product was similar to Dragon Naturally Speaking. During installation, it
adjusts your microphone and speaker volume, followed by a general speech
training session. The software lets you perform two functions. It has a feature
called VoiceCenter to help you navigate and control your desktop and programs
with your voice. You can launch applications, and open and close files using
this. The other function is called SpeakPad, which is similar to the Wordpad in
Windows, where you can dictate text.

The VoiceCenter software worked pretty well after the general
training. We used it to open and close most built-in Windows applications
like Notepad, calculator, etc. To work with the SpeakPad, however, you need to
train the software. There’s an enrollment option for doing this. When you
enroll, you have to dictate 265 lines to the software for it to be completely
familiar with your voice. It takes about 45 minutes to an hour for doing this.
One drawback here was that while you’re reading the text in a sentence, there’s
no indication of how much you’ve read. It’s after you finish the complete
sentence that you get to know whether there was an error in recognizing it. So,
if there’s an error, you have to repeat the entire sentence, as it doesn’t
recognize partial sentences.

After the training, the Text-to-speech recognition rate is
pretty good, though we didn’t find it as accurate as Dragon Naturally
Speaking. While dictating, we found that it worked better if you spoke complete
sentences instead of single words or segments.
It appears to apply grammar rules to whatever you’re saying for creating the
right sentence. You can even play back what you’ve told it to type.
Text-to-speech conversion rate is also pretty good.

ViaVoice has a separate vocabulary manager and expander. You
can train the software to recognize typical words such as names, interjections,
etc, using the Vocabulary Expander. You can view the existing vocabulary list in
the Vocabulary Manager. From here, you can also change the pronunciation for a
particular word.

Overall, the software works well for commanding your PC with
tasks. The text-to-speech conversion, however, needs a little improvement.

Ankur Saxena

at PCQ Labs