Draft 9/25/98 for November, 1998 Law Practice Management Article

JAE Notes March, 99:  To see the PDF file that contains the article as actually published, go back to www.lawtech.com/jimtips/voice.

What follows is the draft of the article as submitted to the editors, with some changes made around the time of publication.  Here is how the following is different from the article as published:

• The following contains links to vendors.
• The following contains some updated information
• The following is longer.  For space reasons, some of the following was cut.
• Style edits were made, so the following doesn't read as well as the one that was edited by a style editor.
• They moved things around, so the order is a little different. 
• I have updated the following with some info that is not in the original article. 

JAE Note:  I have indicated new or updated information that may not have made the magazine article in Red.  . 

"Talk to your Computer -- Practicing Law with Speech Recognition Software"                         

Review of leading speech recognition products, including Dragon NaturallySpeaking, IBM ViaVoice 98, Lernout & Hauspie VoiceXpress, Typhoon and SpeechLaw/Philips continuous speech recognition for Windows 95/98 and Windows NT.

(c) James A. Eidelman and American Bar Association1998
All rights reserved

Eidelman Associates
317 S. Division, Suite 187
Ann Arbor, MI  48104

www.lawtech.com
734-769-1500

Introduction

Whether you currently dictate to a dictating machine ("batch mode") or use a keyboard to draft documents ("interactive mode"), speech recognition software can make you and your secretary more productive.  And you won't have to change the way you work.  I work both ways, and can at times agree with both sides of the debate between attorneys who have strong convictions about the "right" way to work.   Of course, both are "right," and there are flavors of speech recognition software to suit any lawyer's needs. 

For those who now use a dictating machine:

Since I began in practice in 1973, I have practiced law with a dictating machine. Many lawyers who use a dictating machine believe that typing is unproductive.  Their position is that, simply stated, most lawyers can speak faster than they can write or type. For drafting original documents, effective dictation skills will beat a fast typist every time. Moreover, because lawyers spend so much time communicating verbally, dictation is the way their minds work best when being creative. That is, the effort of typing interferes with the smooth and effective flow of billable creativity.

I carry a dictating machine wherever I go, and I use it in the car and just about anywhere else to draft letters, emails, memos, time entries, and even quite lengthy documents.

I haven't entirely given up my dictating machine and transcriptionist yet, but I have been able to go a significant part of the way.  When I am in the office, I dictate to the computer in "interactive mode."  Emails, time entries and short memos I will correct myself.  But, for long documents, now that the software lets you defer correction to a later time, I often use the software in a "batch mode" process.   I save the file, and my secretary can make the corrections, playing back what I said as she works.  When I am away from the computer, I currently use a Norcom dictating machine.  If I am in town, my secretary will "play" the dictation into the computer, and she can make the corrections.  If I am out of town, I will "play" the tape into the computer and transfer the document by email to her for her to correct.  (There are some inconvenient aspects of this, including the need to have separate voice files for interactive and dictating machine use, as well as for the different PCs I use.  In the near future, I hope to also be able to use one of the dictating machines that lets you read a digitally recorded WAV file directly, but as of this writing, I have not tried one.  By the time this article is in print, I will have tried several of these and will provide an update on the magazine's Web page, www.abanet.org/lpm/______) **Joan and Dan fill in.

With that approach, the text can be fed into the computer, automatically converted into text by the software, and corrected by a secretary or lawyer.

Some software packages support "deferred correction," letting a staff member highlight any text that looks strange, and listen to the words that were dictated. This makes it possible for a secretary to work with the text and voice files together while formatting and further edits are made.  In fact, SpeechLaw's implementation of Philips SpeechMagic, a high-end solution for batch processing of dictation, is built around this design. 

See Side Bar, "Special Digital Dictation Units to Capture your Voice."

For those who use a Keyboard:

Lawyers who use a keyboard feel they have a big advantage. 

Seeing your thoughts come alive on the screen aids the creative process. The ability to edit as you create can allow you to do the equivalent of several rounds of dictate-edit-markup-review in one sitting.  I enjoy drafting with a regular word processor, and even more with a creativity enhancer like an outline processor, such as Grandview and Ecco, or the outliner built into Microsoft Word and WordPerfect. 

As a fast typist, I can draft almost as fast at the keyboard as I can dictate, so when I am sitting at my desk or laptop, I use the computer, and when I am walking or in the car, I dictate using a portable dictating machine.

Unfortunately, too many years of heavy typing are now causing wrist pain. The Microsoft Natural Keyboard solves most of my problem, but speech recognition software is even better. You can talk to the computer, and the words appear on the screen. It sure beats typing!

Speech recognition software used interactively, gives you the best of both worlds. You can create text by talking to the computer, not having to type as you create new text. And you get to work with your documents interactively and visually, without the need for your secretary to transcribe a tape.

You can sit at your computer, type some and talk some. You dictate to the computer, and watch the words appear on the screen as you talk, correcting and formatting as you go.

Lawyers who are comfortable with a computer and keyboard will prefer the interactive mode, speaking into a microphone or a headset. You need a very fast computer and decent computer skills, but you don't have to mess with disks or tapes.

How I use speech recognition software:

I have begun using speech recognition software on a regular basis, all day, every day, when I am sitting at my computer.

My work pattern: As the day goes on, I typically have six or more windows open on my desktop, including Outlook (email, calendar, contact manager, task list), time entry, word processing, Web browser and FrontPage.

As I work, I switch back forth between windows. For example, if I receive a phone call, I may type in notes during the call, then draft a word processing document, an email, a time record, and a task or calendar entry.

I have been using a combination computer and telephone head-set, with a button I can press when the phone rings to change modes which I strongly recommend.  This arrangement allows you to avoid having to change headsets when the phone rings.

With the low-end versions of continuous speech recognition software, and on slower machines, you cannot dictate directly into Microsoft Word, Outlook, Timeslips, or FrontPage. Rather, you dictate into a simple "SpeechPad" or "VoicePad" word processor. If you wish, you can format and print right from the simple word processor, although the formatting is less robust than WordPerfect or Microsoft Word. You can also save the document as an RTF (rich text format) file or DOC (Word) file which can be read by most other programs, or you can copy and paste the text into the other program. 

With the earlier versions, the way I worked most of the time was to dictate into the voice window, then cut and paste the text into the time record, email message, Word document, or other application.

Most of the packages now let you dictate into any window, so you can speak right into an email reply, time record, or your favorite word processor.  Microsoft Word users will be happy that just about any of the products has special integration with Word.   WordPerfect users, however, are not left out, as there is full integration with WordPerfect from Dragon and Typhoon.  I have tried all of these packages, and they work quite well, although you need a snappy machine when working with other applications.  

When I'm in applications other than word processing, I talk right into the application, which is particularly helpful when  entering time records. 

Proofreading required:

The speech recognition software makes no typos. However, it makes lots of mistakes, especially at first.  (See below for description of typical mistakes.)  Most of the time, the mistakes don't do any damage. Confusing "in", "an", and "and" generally only affects the readability of the sentence. However, on one occasion the software typed "doesn't" when I said "does" and completely changed the meaning of what I intended.  I did not catch it, and the error would have caused serious problems if a proof-reader had not caught the error.

I find it particularly difficult to proof-read my own work when using speech recognition. When proofreading, one's brain scans the text looking for misspelled words, without reading for content. It takes another person to scan the document looking for words or phrases that do not make sense in context. When I send an urgent reply to an email that someone else hasn't proofread, I sometimes put a note at the beginning that says, "This text was entered using speech recognition software, assuming you prefer a prompt response to perfection. Please excuse any 'funny' or sound-alike words in the text that I may have inadvertently missed."

Impact on productivity:

I liken using speech software to learning to ride a bicycle.  When it isn't going well, you cuss a lot and get nowhere.  When you "get on a roll," using it can feel like coasting with the wind blowing through your hair.  You can crank out letters, emails, and other documents quickly and easily.  Talking is less tiring than typing and being able to get instantaneous turn-around, without needing to wait for transcription, is an important benefit. 

Changing Staff Role: Your Secretary as Editor and Proofreader

The system will make lots of errors, some of which even change the meaning of what you said.  You MUST have someone else proof-read your document!!

Getting Started

Each of the interactive packages is easy to learn, assuming you already know how to use a PC with Windows.  (If you don't, you should get basic PC training before you start.)  Each has a startup procedure, in which you have to go through the following steps: 

1.  Read the simple instructions on how to get started.

2.  Install the software on your machine.

3.  Adjust the audio controls for your microphone and speaker. 

4.  Review on-line introductory documentation and tutorials.  I particularly liked the IBM audio-visual demos that explain things while words appear on the screen and the mouse clicks on various controls. 

5.  "Enroll."  The high end versions allow multiple users to be set up on one machine.  The computer needs to know who you are so that it can work with your vocabulary and voice files.  Each person speaks differently, and the computer needs to get to know you.  Enrollment takes an hour or so.  During the process, you read a short story or two and some explanatory material while the computer instructs you how to speak to it.  When it is done, it takes an hour or so to process your language to learn how you pronounced the words in the enrollment exercise. 

6.  Take it for a spin.  Try out dictating to the computer, watching your words appear on the screen. 

7.  Build the computer's vocabulary in two ways: 

• Use the software's "topic builder" facility to read word processing documents that contain words that you commonly use that are not in the dictionary.  You get a list of words that are not in the dictionary, and you can select the ones you want and even speak them to further train the system. 
• Correct errors as you go, training the computer about the words you did not recognize.  

This process is best done on a weekend, when you have time to experiment without being in the middle of "billable hours" and the phone ringing.  If you play with it, teach it, and learn from it on Saturday and Sunday, by Monday you will begin to be effective in dictating to the computer. 

If you are working in batch mode, it is much easier for the attorney.  The SpeechLaw/Philips software requires very little enrollment, and as the secretaries make corrections, the accuracy will increase.  Like the others, the more you use it, the fewer errors it will make. 

If you are using one of the interactive packages in batch mode, you need to speak the enrollment exercises into the dictating machine, and play them back so that the computer can learn how you speak.  And the accuracy will never quite reach the level of interactive use because the lawyer isn't there to record the pronunciation of problem words.  If the secretary corrects the word without recording, the system doesn't learn as well for the future. 

Training the computer:

You can train the computer to recognize phrases or words you use. For example, I quickly trained the system to learn the names of people and streets that I use regularly.   However, there are still some hard ones when the new words sound like standard words. 

Here is a screen shot of Dragon's Topic Editor screen to add a new word: 

Failures to Communicate:

To get the computer to understand you, even after you have trained it to understand your voice, you must:

• Enunciate very clearly
• Emphasize consonants
• Buy a high-end microphone (see side bar)
• Slow down a little until the system becomes reliable

The computer has trouble with "sound-alike" errors.  It's hard to get mad at the computer for not recognizing mumbling.  But it can be frustrating when you think you are speaking clearly, and it just isn't good enough.    For example, when I said:

I sure look forward to seeing you

The computer heard:

Assure look forward to seen in you

When I repeated the same words with better enunciation, the computer got it right.

Here are some other examples of some "failures to communicate" between the computer and me:

Using a Laptop

Some laptops just don't work well with speech recognition software, even if they otherwise have enough power.  There can be two reasons:  (1) inadequate sound card, and (2) the circuitry is jammed so close together that there can be interference.   For example, the Dell Inspiron is an inferior computer to the Dell Latitude CP, but the Inspiron worked better for me with speech recognition.  If you are planning to use speech recognition software and have not yet bought your laptop, experiment first, and don't forget to experiment with two or three different headsets.  The results can vary widely.  Many laptops just don't work well.  One brand that does is the Micron TransPort line.  Micron has worked closely with Dragon, and Dragon recommends the Micron products.  Similarly, I assume that the high-end IBM ThinkPads work with IBM's ViaVoice, but I have not tried them.

On the Road

Noisy background? I am happy to report that I was successfully able to dictate on the train from D.C. to New York, in spite of an extraordinary amount of background noise from the road, the fan and voices of my fellow travelers. I was even able to dictate directly into my laptop computer in the car while driving.

Supporting Speech Recognition in a Law Firm:

I do not recommend simply buying the software and delivering it to the attorneys to pick up on their own.  The average attorney will simply get frustrated and fail to achieve the benefits.  Only the "techies" will succeed. 

To be successful, most firms should:

• Set up a pilot group of patient lawyers to try it out.
• Have the "techies" use the PC directly.  Team them up with a paralegal or associate to proofread their briefs, pleadings and contracts, and a secretary can make corrections on simple letters.  Only use software that lets the secretary or proofreader listen to what was dictated. 
• Have a pilot group of lawyers who already use dictating machines enter text using a dictating machine, such as the Norcom/Dragon or IBM dictating machine combination. 
• When you roll the system out to the firm, be ready with personal trainers and floor support. 
• Require everything to be proofread carefully. 

Attorneys need to take care not to let the proofreading and correction time kill their productivity.  If I can dictate or type an agreement in an hour, it may take me an hour and a half to dictate it into the computer and correct it.  I have saved my secretary the trouble of transcribing the tape, but I have lost productive time.   Some attorneys will have a smooth ride, and after they have been using the system interactively for a few weeks and suffered some loss of productivity, their productivity will rise again to where it was before.  Others will not.  For them, most of the time, they should work in batch mode, letting the software serve as a tool to improve the productivity of the secretaries.  Each firm will need to experiment to see what works for them. 

The Vendors and the Market:

Speech recognition software packages work in one of two modes:

• Discreet: The early packages used what is called "discreet" speech recognition. Early versions included Dragon Dictate and IBM VoiceType. These packages work well, but you cannot dictate in a conversational tone. Rather, you-must-speak-like-a-ro-bot, separating every word and syllable with slight pause. Most of us would rather use a dictating machine or type than have to dictate that way.

• Continuous: The new packages use "continuous" speech recognition. As long as you enunciate very clearly, you can dictate to the computer much as you would dictate to a transcriptionist. This is what most lawyers have been waiting for. All of the packages described in this article are "continuous." 

The market for continuous speech recognition is experiencing roller-coaster-ride excitement, with dramatic changes every few months. New features, new versions, and price changes abound. 

I will leave it to the trade press to describe the current features of each package. Journals like PC Magazine and Windows Magazine are in a much better position to present blow-by-blow comparisons of the new versions of the products as they are released.   (For example, see the articles referred to in the side bar, "For more information...")

The following is a general description of the offerings from each of the vendors:

Products that work interactively on your own PC:

Dragon NaturallySpeaking

Dragon Dictate, a "discreet speech" product, was the first PC-based speech product on the market, and was first to market in the summer of 1997 with a "continuous speech" product.  It has an excellent interface.  In August, 1998 Dragon began shipping new, enhanced versions of the software.

Dragon NaturallySpeaking won more than 20 major awards and was named "Best of the Year" or "Editor's Choice" by just about every reviewer, from Time Magazine to PC Magazine.  Its interface is excellent, especially for making corrections and "training" the system about how you pronounce words it didn't understand.  The "vocabulary builder" even lets you feed it word processing documents to add words that aren't already in the dictionary. 

The WordPerfect Legal Suite 8.0 now comes with a free copy of Dragon NaturallySpeaking Personal Edition, the basic and least expensive version.  (You'll want to upgrade because the personal/standard edition doesn't include the Mobile module to let your secretary convert and hear your sound files.) 

The "Preferred" and "Professional" versions add a number of features, including, most importantly, the ability to talk right into your word processor, the Mobile module transcription of WAV files, and the ability to create voice macros.  The single most important feature of the new software is the "NaturallyMobile" module.  The software can directly transcribe a standard Microsoft WAV sound file, which can come from text dictated directly to the PC using the "Sound Recorder" software (dictated directly or through a wire from a Norcom hand-held unit, or which can come from a WAV file captured on a digital dictation machine.  Your secretary can listen to the sound file that goes with the text as she corrects during that session, although the sound file is erased once the session is closed.  You cannot yet save the sound file for later correction.   (Note that the WAV files should be in the following format:  PCM, sampling at 11,025 Hz, 16 bit, Mono (21 kb/second)). 

Dragon also offers a Dragon NaturallySpeaking Legal Suite specifically designed for lawyers. It is a combination of the Professional version plus an extensive legal vocabulary of over 240,000 words, which includes court names, case history and weight of authority, Latin and French phrases, litigation terms, and abbreviations. You say the term the way you would say it in conversation (for example, "New York second") and the software interprets it and types it correctly ("N.Y. 2nd"). In addition to the preexisting vocabulary, you can also create your own vocabulary with terms from your legal specialty.  Dragon NaturallySpeaking Legal Suite does not come with predesigned macro templates, you can create your own. 

Finally, if your computer has extra RAM, the new version also supports "BestMatch technology," a module that increases performance and accuracy, and also increases the size of the vocabulary, and is strongly recommended. 

IBM ViaVoice 98

IBM has been working in speech recognition for decades, and originally developed continuous speech recognition for minicomputers.  The first ViaVoice 98 product for PCs was released in 1997.  Most users, including myself, didn't like its interface or usability nearly as well as Dragon's, but it set a standard in price (very low!) and in one important feature -- the ability to save the dictation with the text of the converted document so that dictation could be done in batch mode, with correction deferred until later, to be done by a secretary. 

The new ViaVoice 98, released in July, 1998, is the latest addition to IBM’s recognition software family. At $149, its "Executive" version is by far the least expensive full-featured product.   With ViaVoice 98 you can open files and e-mail, format text, dictate documents and e-mail directly into most Windows 95 and Windows NT applications, although the menu commands don't work in Windows NT. This software package has the macro capability mentioned above, so you can eliminate repetitive typing by calling in standard paragraphs, phrases, form letters, addresses and so forth using simple verbal commands. For example, by merely dictating "thank you," the entire body of a standard thank you letter will appear on your screen. Another example would be to print your document by simply uttering "print."

ViaVoice 98 features some context recognition, which improves recognition of sound-alike words like words and phrases like "there, " "their, " and "they’re."

ViaVoice 98 also offers a legal dictionary which retails at $149. 

Topics for legal specialties, such as worker's compensation and corporate law are now, or will soon be, available for purchase on IBM's Web site.  They have been developed by SpeechLaw, and offer a combination of specialized vocabulary and language model. 

This version offers a much-improved interface, improved accuracy, and other features.   I have been using Dragon for a year, so of course, it feels more natural to me.   I am beginning to use the IBM product, and as I learn it and get used to it, I like it.  The 98 version is a complete makeover, so anyone who rejected the IBM project based on the earlier version should take another look. 

Olympus offers a combination product, the D1000, that includes a digital dictating machine with a customized version of IBM ViaVoice 98 which looks like a terrific combination, although recognition in the initial version is not quite as good as expected.  Dictaphone and Norcom now also offer versions of IBM ViaVoice that are bundled with their hardware. 

Special versions that directly read a WAV file are or will be available from makers of dictating equipment, such as Olympus and Dictaphone, but as of this writing, the off-the-shelf versions of ViaVoice from IBM do not yet match Dragon NaturallyMobile's ability to do so.  IBM ViaVoice 98 does save dictation files so that a secretary can listen to the dictation as she corrects.

Lernout & Hauspie VoiceXpress

The new kid on the block is L&H VoiceXpress Plus.   L&H began in language translation software, then acquired the product Kurzweil had invented. Then Microsoft invested over $40 million in the company for 8% of the stock. The first L&H package was not designed to dictate the text into a document, but rather to control Windows and format Microsoft Word documents with voice commands. Their first continuous speech product, released May, 1998, stands out in its ease of controlling Microsoft Word with plain English voice commands and the ability to dictate numbers and letters in a natural way.

By the time this article is in print, L&H expects to have announced new versions of its product line that, like Dragon and IBM, have four levels of functionality: Standard, Advanced, Professional, and Legal. 

The industry has great expectations for this product line, as Microsoft moves toward adding speech commands as part of the operating system in the future.

Philips FreeSpeech

Like IBM, Philips, a Belgian company, has been working in speech recognition for two decades.  Until now, Philips has only marketed high-end solutions through third parties targeted at the professional market (e.g. SpeechLaw), and Philips has not   offered a consumer product.  New to the market is Philips' FreeSpeech, which you can download and test for free from  www.freespeech98.com, assuming you have a high quality microphone.  It has some very nice features, such as its patented dictation playback with synchronous highlighting, ability to work with Microsoft Office and other applications, and ability to improve performance and accuracy by learning from correction of recognition errors. 

Typhoon

Typhoon produces a specialized adaptation and interface with the IBM ViaVoice 98 "engine" to optimize its performance in the major word processors (Word 97 and WordPerfect 7 and 8), plus Novell's GroupWise and soon MS Outlook.  It has produced 500 commands for editing text and gives a complete and natural experience for using speech to create documents. Typhoon's founder is a lawyer who has taken special pains to produce a product that is optimized for the professional user.  It is available in a version that allows the user to save the original voice dictation so that a transcriptionist can listen to the original session while making corrections to the draft, then delete the voice file which is memory-intensive.

Typhoon representatives say that the products will be improved later this year with the addition of artificial intelligence for further accuracy and speed enhancement.

Batch Mode, converting sound files into text on a server:

SpeechLaw (uses Philips SpeechMagic)

Philips created the first continuous speech recognition product for PCs, and through SpeechLaw, the creator of the legal "ConTexts," began marketing to lawyers at the ABA Techshow in 1997.   This is not a shrink-wrapped product, but rather one that is sold for network integration, often with training and integration services, at a price of $1,500 to $2,500 per attorney (purchase price) or $125 per attorney per month (after setup) on a subscription basis.  Consulting services are extra. 

SpeechLaw's software was designed so that lawyers who dictate with a dictating machine can just keep doing what they have always been doing.  They don't even need to have a PC, and if they do, it doesn't have to be a powerful one.  The recorded sound files are transferred to an speech server running on a Windows NT machine, converted into Word or WordPerfect documents, and routed to the secretary or back to a lawyer.  A secretary can use a foot pedal/headset combination to listen to the attorney speaking while the words being spoken are highlighted on the screen.  She can stop, back up, and make corrections, and the corrections will be used by the system to improve accuracy in the future. 

The Philips software cannot work without a "ConText," which is a combination of a specialized vocabulary and voice model.  SpeechLaw has set up special ConTexts for a wide variety of areas of practice, including many litigation specialties, real estate, corporate law, and labor law. 

The software also has database compatibility through Microsoft's ODBC (open database connectivity) standard, so it can ready SQL and Access databases.  If the integration work has been done and the system connected to a case management or other database, an attorney can dictate the variable names in the text.   When the sound files are converted, the package can look up the names and other information and replace the variables with the text from the data file. 

The vendor reports that, over a period of time, its clients have achieved dramatic improvements in lawyer-to-secretary ratios. 

Conclusion:

The Dragon, IBM and the other interactive products are extraordinary values. So is SpeechLaw, when viewed in light of the value of an attorney's time and a secretary's efficiency. We strongly recommend putting the money into a fast computer and good microphone, and putting the time into learning and training the software. And we are certain that in the future, each of these packages will continue to become easier to use, more accurate and more robust.  But you don't need to wait. Now is the time.

SIDE BARS

Dictation Units to Capture Your Voice: As described in the article, Philips and the "high end" versions of Dragon and IBM also support batch entry of voice that was recorded on a special digital device.  Dragon recommends the Norcom 2500. It currently sells for about $500, and has a special adapter that lets a tape play into the PC.   Dragon's new NaturallyMobile module also will directly read Windows WAV files that are captured digitally.  Philips hardware and software work together.  The Olympus D1000 digital dictating machine records a sound file, and, as with the Philips, the customized version of the IBM software recognizes the digital WAV sound file directly.  The Norcom 2500 with SRC-1 Speech Recognition Coupler (right) is a high-end mini-cassette recorder that functions just like any other dictation machine.  Norcom Electronics has worked with Dragon Systems and IBM to optimize it for use with their speech recognition software.  The SRC-1 is necessary to change the line-out signal from the dictation machine to be compatible with the microphone-in jack on your computer when playing your recordings back for transcription.  With the SRC-1, the Norcom 2500 sells for around $500, but you may find a special offer.  We have tried it, and we like it.  (http://www.norcom-electronics.com/) Still believed to be the most accurate of all.  Since the article, Norcom has introduced the SpeechPort, a combination mike and dictating machine that won the SpeechTEK Peak Performance Award.   They ship it bundled with IBM's ViaVoice, or you can use it with Dragon NaturallySpeaking.   Voice-It unit marketed by Dragon as Dragon NaturallyMobile.    Marketed with Dragon NaturallySpeaking, and optionally with ACT contact manager integrated with Dragon.  For more info, see http://www.dragonsys.com/frameset/product-frame.html and http://www.voiceit.com/vtrline.html. Sony markets MZ-R50, MZ-R30, and MZ-B3 mini-disc recording Walkman.  In theory, offers the advantage of digital recording with much larger storage than PC Cards.   The earlier models we tested didn't have the dictating machine controls lawyers need, and we haven't seen or tested these yet.  Stay tuned for our comments.  Olympus markets a digital dictating machine that saves the sound files on a PCMCIA (AKA "PC Card") memory card.  They bundle it with a customized version of IBM ViaVoice 98.  The benefit is that, unlike the Norcom or Sony, you don't need to play the dictation through a wire back into the computer.  Rather, the software just reads the sound file and uses it.  List price is $349, with an expected "street price" of $299.00.  One of the competitors claims that the accuracy is not high enough, but we look forward to trying it.  We also understand that the Olympus works great with the SpeechLaw software.  Philips began in the dictating equipment business, and that heritage shows. An attorney who is comfortable with traditional dictation will feel right at home with either of Philips' two hand-held units: The SpeechMike attaches to the PC, and has the look, feel and function of a desk-top dictating machine. You will find a comfortable combination of buttons for fast forward, rewind, record, listen, stop and pause.  Sound files are compatible with Norcom portable unit, so you only have to have one set of speech files if you use this combination on and off the PC.   They also market a rather pricey ($2,000 at the moment) portable unit that captures almost an hour of speech to a PC Card memory card. That card can then be plugged into any PC Card slot to have the voice files processed by the speech recognition software.   Dictaphone's Boomerang, like the Philips SpeechMike, will be a natural fit for lawyers who are used to working with a desk-top dictating machine with hand-held mike.  It has the normal buttons and slide controls you can control with your thumb, only because it is digital, not only can you go forward and back and insert spoken text in the middle of other spoken text.  It saves the speech as a standard Windows WAV file.  They also make a companion unit for the secretary, with headset and foot pedal, for manual transcription.  In November, 1998 Dictaphone is expected to release a customized version of IBM ViaVoice 98 that is bundled with the Boomerang products.  They will also market the Quest, a high-end portable digital dictation unit with some palm-computer features.  As of March, 1999, the portable products are not certified for speech quality (Quest and Walkabout Tour).  For more information, see http://www.dictaphone.com/ordering/shop/boomerang/boom.htm

Integrated dictation hardware/software solutions for batch dictation and deferred transcription:

*Pricing may vary with licensing fees

Software Pricing Information

*Pricing may vary with licensing fees

For More Information:

Other reviews and articles:

Links to Software Publishers:

Some Links to Pages of Vendors, Consultants, etc.

VXI:  http://www.vxicorp.com

Andrea:  http://www.andreaelectronics.com

Dictaphone: http://www.dictaphone.com/products/boom/boom20_index.htm

Norcom Electronics: http://www.norcom-electronics.com/

Voice Recognition Systems, Inc. Home Page: http://www.iglou.com/vrsky/

21st Century Eloquence: www.continuous-speech.com

Speech Recognition at Techshow 99

If you want to learn more about Speech Recognition Software first-hand, come to the ABA Techshow 99, which will be held in Chicago, March 18-20, 1999.  This three-ring circus on law office technology will include presentations about speech recognition and the opportunity to visit vendors on the exhibit floor and see speech recognition in action.  For more information, point your Web browser to www.abanet.org/techshow, or call 312-988-5___ (????? Joan -- Please fill in). 

Frequently Asked Questions:

1) How do "continuous speech" products differ from "discreet" speech products?

"Continuous speech" products allow you to speak in a normal conversational way to your computer, whereas "discreet" speech products require you to pause between words, like a robot. 

2) Are these products ready for general law office use? Which lawyers should use them?

These products have reached a level of power and flexibility that make them more than adequate for general law office use.   Lawyers should note however that there is a learning curve involved with integration of these products into daily work.  With the interactive products, the lawyer will need to be computer-literate and patient at first

3) How do the products compare?

SpeechLaw's implementation of Philips SpeechMagic is a network-based system designed for batch-mode use.  The software is optimized for lawyers to dictate into a dictating machine, as they traditionally have, and to automate the secretary's job.  IBM and Dragon are designed for interactive use, but now both support deferred correction by a secretary for batch processing (using two different methods).  Microsoft is expected to build speech recognition into future products, and has invested in products from Lernout and Hauspie.  The Lernout and Hauspie VoiceXpress Plus has the best voice control of Microsoft Word and Windows. 

4) Do you have to talk to the computer, or can you use a dictating machine?

It is possible to use a special dictating machine to record your dictation and then play the recording back through the "Professional" or "Executive" versions of the software.  Norcom offers an analog unit optimized for this market, and Sony offers a digital minidisc unit (with woeful controls). Olympus allows you to record the speech to a WAV file saved on a PC Card.  Their customized version of the IBM ViaVoice 98 software eliminates the need for the tape to be played into the computer.  Philips has a similar solution, although it is much more expensive.  Philips and Dictaphone offer products that are like a desktop dictating machine.  Dragon's NaturallyMobile will also transcribe a digitally recorded WAVE file.

5) How does this software change staffing needs?

Secretaries will spend much less time transcribing tapes, but more editing by secretaries or other staff is required.  The net result will be a lower "head count" of non-lawyers and better use of staff time.  And setting up voice macros will make an even bigger impact. 

6) What skills does a lawyer need to use these packages?

To competently  use the interactive speech recognition packages, a lawyer must have (a) basic skills in using a PC with Windows and Word or WordPerfect and (b) some time and patience.  Good dictation skills don't hurt either...  In batch mode, great dictation skills are required, but a lawyer doesn't need to know how to use a computer (assuming a secretary does the correction). 

7) Can I talk right into my word processor and other programs?

Yes.  All of the packages offer a basic version that allows you to talk into their own mini-featured word processing package, and a higher-level version that allows you to talk into other programs.  The "SpeechPad" or "VoicePad" programs are better for long input on a machine in the Pentium 200 range for speed reasons, but with a very fast 300 mh Pentium II, it works fine to dictate right into other applications.  

8) Can I control the computer with voice commands?

Yes, although you need to know the commands, and many users like to use the keyboard and mouse for many of their commands.   The basic voice-related commands are included even in the low-end versions.   The "discreet" programs have always supported this.  I prefer to navigate and correct with the keyboard and mouse. 

9) How much hardware do I need?

You won't notice much delay with a Pentium II or equivalent.  A Pentium MMX 200 is fast enough for the user to be just barely satisfied.  A laptop with 233 mhz "Tillamook" Pentium MMX processor is what I have been using most of the time.  I have 64 mb of RAM.  Allow 32 mb for the voice products on top of what you need for Windows, Word, Outlook, etc. The new version of Dragon has higher recognition and some additional features that you can use if you have an additional 32 mb.  And don't forget that NT takes 16 mb extra.  If you are running all of the features of Dragon on an NT machine, get 128 mb. 

9) Can my secretary listen to what I dictated as she corrects the errors?

IBM ViaVoice 98:  Yes.   Lernout & Hauspie VoiceXpress Plus:  No.  Dragon NaturallySpeaking Preferred and Professional:  No, other than once if transcribing a WAV file.  SpeechLaw/Philips: Yes.)  

Speech Recognition Companies

Andrea Electronics Corporation
Long Island City, NY
(800) 442-7787
http://www.andreaelectronics.com/

Dragon Systems Inc.
Newton, MA
(617)332-9575
www.dragonsys.com

IBM Speech Systems
West Palm Beach, FL
(800)825-5263
www.software.ibm.com/is/voicetype

Typhoon for Word and WordPerfect
Santa Barbara, CA
800-933-6520
www.typhoon.com

Lernout & Hauspie
Burlington, MA
(617)238-0986
www.lhs.com

Philips Speech Processing
Atlanta, GA
(770)821-2400
www.speech.be.philips.com

SpeechLaw
Mission Viejo, CA
949-837-8272
www.speechlaw.com

Home to www.lawtech.com

*******************JAE NOTE:

The following is a copy of the software pricing chart with 2 extra columns in it showing whether software can transcribe a WAV file and whether software can save dictation with text for deferred correction -- 2 important features. 

***************************************

Software Pricing Information

*Pricing may vary with licensing fees

Other Links:

The Good Dragon does Speech Recognition very well, http://www.perigee.net/~gttracy/speech.html

Forums from 21st Century Eloquence: http://www.voicerecognition.com/forum/index.html