Compare & Contrast

Law Technology News (Volume 6, Issue 10, p. 38, col. 1)
October 1999

Voice Recognition: The Three Legal Packages

by Wayne Spivak

EVER SINCE the '60s and the introduction of artificial intelligence, we have dreamed of the day when we would be able to speak to our computers. Actually, we've always been able to speak to them, yell at them, scold and coax them, but truthfully, they never understood.

Today, 30 years after the inception of the Internet and six Star Trek movies later, we are beginning to see the ability to talk to our computers, and their ability to understand us.

At a recent press conference, Leonard Kleinrock, Vint Cerf, Robert Kahn and Lawrence Roberts -- all considered the "fathers of the Internet"-- spoke about the 30th anniversary of their wildly growing offspring. Kleinrock postulated that in the future, "The walls will be alive with technology. We will be able to move into an environment and the environment will know we have just arrived."

However, Roberts also reminded us that interaction between ourselves and our computers hasn't quite progressed as fast as it should have during the last 15 years.

So, where are we in this technological niche? Today, several vendors are producing software that permits the end-user to talk to the computer. These software packages either permit words to be reproduced in a word processor, or interpreted as computer commands, such as "print document." So instead of sitting at my desk typing this article, I could have dictated the article into a microphone, and been presented with a finished, polished document, or so the theory goes.

Dragon System's NaturallySpeaking Legal Suite
(Circle no. 345)		 Lernout & Hauspie's VoiceXpress for Legal
(Circle no. 346)		 IBM's Via Voice Legal
(Circle no. 344)
MINIMUM REQUIREMENTS: 133 MHz Pentium Processor

IBM-compatible PC with Windows 95/NT 4.0

MEMORY: 32 MB RAM for Windows 95, 48 MB for Windows NT

HARD DISK: 120 MB, 40 MB additional to save recorded speech, 15 MB additional for Text-to-Speech

CD ROM drive

VOCABULARY: 240,000 word built-in legal vocabulary, including hyphenation and abbreviation

SPEED: 160 words per minute.

SPECIAL FEATURES:

Vocabulary Builder automatically scans hard disk and its documents to recognize user writing style, most-used words and specialty vocabulary. Can also add words manually.

Macro-enabled, creates macros for boilerplate text, and commands. Command Wizard enables Naturally Speaking to integrate with other Windows programs.

Dictation Playback; Text-to-Speech

 MINIMUM REQUIREMENTS: Pentium 166 MHz processor with MMX (Pentium II recommended)

IBM-compatible PC with Windows 95/98 or Windows NT 4.0 with Service Pack 3 or later

MEMORY: 64 MB RAM

HARD DISK: 190 MB

16 bit Windows-compatible soundboard

CD-ROM drive

VOCABULARY: Built-in legal specific vocabulary, can be customized by up 64,000 words

SPEED: 140 words per minute

SPECIAL FEATURES:

Text-to-Speech playback

Vocabulary Extender integrates user's writing style with program for most used words and specialty vocabulary

Talk-and-Go; users can dictate to an Olympus DS-150 digital voice recorder for playback to the computer later.

Macro-enabled; can activate macros both via keystroke and voice

MINIMUM REQUIREMENTS: Windows 95/98/NT, Pentium 233 MHz processor and 256k L2 cache including AMD-K6 233 MHz with 256K l2 cache, 48 MB RAM, 290 MB hard disk space for the Web version, 310 MB for the Pro version, Windows 95/98 compatible 16-bit soundcard, Internet Explorer 4.0 or higher (Web version).

VOCABULARY: 25,000 legal terms, both criminal and civil,

SPEED: 140-160 words per minute

SPECIAL FEATURES:

Text-to-Speech playback.

Automatic "Analyze My Voice Feature" - learns and remembers how you speak

Macro-enabled; integrates with other Windows programs.

Voice Mouse for hands free mouse functions and navigation of desktop.

Speech-enabed web browsing for America Online 4.0 or later, Netscape Communicator 4.5 or later and Internet Explorer 4.01 or later

Macintosh version slated for release later this year

Linux Software developers kit is available

Let's not discuss a complete realignment of my mindset. I don't really like writing pen on paper, and never, ever want to return to the days of the typewriter. I've never dictated to, nor even used a dictation machine. So to use a voice-activated interface with my computer would take some training on my part.

This brings us to one of the most important issues surrounding voice recognition products: they require training! Lots of training. Which means you must decide from the outset to invest the necessary time into training your software to understand your speech patterns.

Even then, after you've invested the time, you'll only have a product which, today, is by all accounts only about 90 to 95 percent accurate. One of the major complaints made about voice recognition products is their error rate.

Error rate is defined as the percentage of words the product misinterprets. It is a tedious process to add dictionary words, and without them, the odds for any voice recognition product to err on industry or legal specific words are much greater.

Many of the major players in the voice recognition market now offer legal specific editions. Why purchase the legal specific edition? These special editions contain larger dictionaries, and other special integrated links to increase your productivity.

An example provided by Dragon Systems is the term, "New York second district." When spoken into their product, "N.Y. 2nd" will appear. This attribute increases productivity, and decreases the error rate. Without the legal specific edition, you'd probably have to re-edit the document.

In addition, many vendors are making products that link with other software products. Examples are word processors, time & billing products, and text to speech software. Even a Palm Pilot version has been announced.

How to select the correct voice recognition software? Look at the system requirement vis a vis your computer system (all vendors require Pentium processors and some type of sound card). Look at the feature list of the software product. Compare this feature list to your work habits and needs. No two people work alike, nor are any of these products 100 percent identical.

Finally, be prepared to spend a considerable amount of time with your voice recognition software! Your investment will pay off, by increasing your productivity.

Wayne Spivak is president of Bellmore, N.Y.'s SBA* Consulting.