ViaVoice

Voice recognition systems have been seductive computer add-ons for a long time. Software companies have delivered the software with the promise of freeing your hands of the keyboard and making document creation a more pleasant experience. But the universal question is whether these packages work.

My experience has been that voice recognition packages work, but each package recognizes the spoken word with varying accuracy. I've found that the higher the accuracy rate, the more time I spend uninstalling the tool from my machine. Perhaps this situation is about to change.

IBM has progressed significantly in voice recognition software with the release of its new product, ViaVoice. ViaVoice is a Windows NT and Windows 95 add-on kit that brings highly accurate speech-to-text capabilities to your word processor.

The software first learns your particular accent and speech patterns and then stores this information for later use in recognizing your speech. This process is called enrollment. Although enrollment is not necessary, it offers a few distinct advantages. Enrolling your voice into ViaVoice increases the accuracy of the speech-to-text conversion mechanism. Enrollment also lets you retain your particular voice profile for later use. This feature is incredibly useful if more than one person will be using ViaVoice. Instead of using someone else's voice profile or repeating the enrollment process, you can simply load your voice profile and start speaking.

Installing the software is slow because of the enrollment process, which takes at least 30 minutes to complete the basics. During enrollment, ViaVoice guides you through a series of phrases and sentences. This process lets the software build your individual voice profile. Screen 1 shows one of the readings I had to perform.

After you complete this stage, you can either use the software or continue customizing it. However, keep in mind that while you use the software, you'll have the opportunity to perform corrective updates, which over time, further increase its accuracy.

ViaVoice has several useful components that ship with the package. SpeakPad is an optimized speech recognition word processor that lets you dictate, transfer, and paste text into any other cut-and-paste-compliant software. The Vocabulary Expander lets you add your own words (up to an additional 42,000) to the built-in 22,000 word base vocabulary. Outloud is a multifaceted text-to-speech applet that can read text out loud, as the name implies. Outloud can also function outside of ViaVoice so that, for example, you can have Outloud read you your email.

In comparison with other products, ViaVoice lets you speak more naturally without hesitating between words. You can't spout off like the world's fastest talker and expect ViaVoice to catch everything you say, but you can speak at a reasonable pace while retaining a high level of accuracy in the translation to text.

Accuracy in translation depends on sheer horsepower because ViaVoice matches spoken words to a digital vocabulary on the fly. To keep up with the pace of your speech, ViaVoice needs enough horsepower to search the vocabulary in realtime without causing you to pause in your speech. My 166MHz Pentium workstation with 48MB of RAM was barely adequate to keep pace with my ordinary speech rate. However, using 200MHz systems or faster should yield very favorable results. IBM's specs state that ViaVoice can translate a maximum of 125 words per minute.

ViaVoice includes a headset with a built-in microphone. The software is available in seven languages: US English, UK English, French, German, Italian, Spanish, and Chinese.

Using ViaVoice was a good experience in comparison with using other voice recognition software. The software is highly accurate after you've trained it to understand your speech patterns. Anybody needing to get their hands off the keyboard will want to try ViaVoice.

End of Article