Home
Alan J. Steinberg
Les A. Steinberg
Articles
How to Find Us

LIVE NOTE
VOICE TYPE RECOGNITION

This is what we've all been waiting for!

Let me tell you about VoiceType recognition programs.

When I was about nine years old, I was watching Tom Corbett and the space cadets. I remember a certain scene in that movie where a typewriter was shown typing out correspondence while the Commander of the Space Academy dictated into a microphone. I was astounded by the picture only to be very disappointed by learning that such state of the art equipment was not available. The equipment was not to be available for approximately fifty years to the general public. I am dictating this article using VoiceType recognition software by IBM. The particular software that I am using is Simply Speaking Gold. This version of IBM Voice Recognition software has superseded "VoiceType 3.0" and appears to be a substantially improved product then the former version.

I have also learned that the software is sound card dependent as to its efficiency. For example, before I installed the Voice Recognition software on my NT 4.O workstation, I attempted to utilize the software on a Gateway 2100 solo notebook computer using an Ensonic sound card. The results were much more disappointing using the notebook than the workstation. The Gateway 2100 solo is not a "Wimpy" machine. It runs on a 150 MHZ processor Pentium with 40 meg of ram. However, the NT 4.0. Workstation is certainly a more robust machine running at 266 MHZ using a Pentium Pro II processor with 200 meg of ram. Based on my experience, I would suggest that the user of VoiceType recognition software be prepared to run the software on the most robust workstation one can afford.

I dictate into VoiceType recognition software by wearing a headset. A training module is available where the user reads about 1 hours' worth of words in response to a "bouncing ball" parade of words passing before the user's eyes. This is called "the enrollment". The user repeats the words, and if "understood" by the software, the bouncing ball continues and more and more words flash by the user's eyes. The manufacturers of VoiceType recognition software claim that the accuracy rate will reach approximately 90% to 95% "right out of the box". I did not enjoy that experience. I found that the recognition rate was probably in the 65% area. Certainly, at 65%, one cannot get terribly excited about VoiceType dictation. Nevertheless, after spending some time training the voice modules, the recognition rate has now reached, in my estimation, 80% to 85% [and with Typhoon's add on the accuracy is even higher). When combined with good secretarial assistance, the use of VoiceType dictation for drafting, along with strong secretarial services for completion of the work, the combination make a profitable duo, of that I am certain.

How does VoiceType Dictation work??

Voice dictation commences with a command "Start dictation". Each word, at this time, must be spoken discreetly, (this portion of this article pertained to Via Voice, before the Via Voice Gold upgrade was issued). With Via Voice Gold (the upgrade), the user no longer has to speak "discreetly", and may speak up to 200 words per minute, in continuous speech, but the user must still speak distinctly.

I have witnessed demonstrations of "Continuous voice recognition" by Dragon Dictate and I have been impressed its efficiency. On the other hand, when I attended the ABA Legal Tech show in April, 1997, I also was impressed by demonstration of VoiceType 3.O. Again, with the 1998 ABA Tech show, Voice Type Recognition programs attracted many attendees.

Unfortunately, the reality of the use of the software in the demonstration bore few, if any, similarities to "real life", that is until, in my opinion, the release of the Typhoon add on.

How does the program work?

After dictating the desired words to the program and the user is satisfied with the content of same, a "Correction" process is undertaken. By giving the software proper commands - "Begin Correction or Correct Word" [depending upon which version the user is using, ten words at a time are underlined or placed in a dialog box for the user to view. The user then speaks (or with a mouse "selects") the word needing correction and a list of suggested "substitute words" pops up in the user's modifiable dialog box which allows the user to set the number of suggested words to correct the writing.

I have my VoiceType dictation pop up set for 15 words. If the correct word is not among the 15 suggested words, then I type in the correct word in a small space in the dialog box, and hit enter. Often the program will seek clarification as to whether the word I typed is to be added to its "Dictionary". Alternatively, the program will ask whether the words I have typed are to be considered one word. This inquiry occurs when I type into the dialogue box two words that were necessary for correction.

One of the things a user can do in order to make VoiceType more efficient is the use of the VoiceType analyzer. The analyzer process is a separate program within the VoiceType recognition software suite, whereby the user takes a series of documents represented by path and file name and imports the documents into the analyzer. The analyzer is then told to "analyze" the documents, and the software digests the words from your previous files and builds a vocabulary and a "Style". This is something similar to the "enrollment process" but uses your documents instead of your voice, in order to show the program your style of dictation.

Recently, I purchased from the Typhoon Company a "conduit" to WordPerfect 8. This conduit program which costs two hundred dollars, integrates with WordPerfect 8. ViaVoice Gold does not integrate with WordPerfect at this time. It only works with its own "SpeakPad" program (a minimum dictation program) and Microsoft's Word, word processing program. The integration factor of Via Voice Gold with WP 8 substantially improved the performance of VoiceType dictation.

The Typhoon conduit enables me to command WordPerfect verbally as well as do the Voice Dictation. I would estimate that the accuracy of the program approaches 90 percent after a substantial enrollment was made on my behalf. Enrollment is the process by which the user reads a "story" into the computer, so that the users voice is "understood" by the software. I would not believe for a moment advertising claims that state that VoiceType dictation operates right out of the shrink wrap at 95 percent without enrollment. It has just not been my experience.

So to summarize, let me say again, that the key to satisfactory VoiceType dictation lies in the user's use of a very powerful computer. And the user must speak clearly at all times. In my opinion, attorneys with proper equipment and patience can use Voice Recognition Software right now to alleviate some of the burden that they had to have sustained in the past, by dictating and watching their words appear on screen at the time of the creation of the document. Having a secretary or Legal Assistant to do the formatting and "cleaning up" is, in my opinion, essential and together, the production of "words to paper" can be enhanced significantly.

©1998

Designed by Binaris Consulting