"CAN YOU HEAR ME?--
DragonDictate for Windows Minces Words for your Office"
An edited version of this article was originally published in the Computer Counselor column of the May 1995 issue of the Los Angeles Lawyer Magazine.

Comments are generalizations and may not apply to every situation. It is advisable to consult with a competent professional before relying on any written commentary. Paul D. Supnik is a member of the State Bar of California and not of any other state or country. The material set forth herein is not for the purpose of soliciting any engagements where to do so would be to offend or violate the professional standards of any other state, country or bar. This web page Copyright 1995, 1996 by Paul D. Supnik

Voice recognition programs "hear" words spoken into a microphone and translate them to typed words which appear on the computer screen. Most lawyers have probably heard at least something about voice recognition programs. This article reviews the practical aspects of using the voice recognition program, DragonDictate for Windows, in a law office environment.

There should be obvious advantages to be able to dictate into a microphone and have text appear immediately on a computer screen. One practical benefit would be to eliminate some tasks probably performed by a secretary. That may or may not result in a cost savings depending on the relative use of attorneys time to dictate into a microphone rather than a dictating machine. Another benefit would be the ability to see words immediately on the computer screen. A very significant benefit would be to be able to communicate thoughts directly from the attorneys mind to the screen while maintaining a rapid flow of ideas. This could be perhaps one of the most important advantages for using voice recognition programs. Can these potential benefits be realistically attained today?

There are serious limitations on using voice recognition programs in a law office to fulfill these objectives. The limitations include training time; the program most be trained to recognize the manner in which words are spoken by the lawyer, and the lawyer must be trained to understand how to effectively use the program. A dictating machine frees the lawyer from placing himself or herself in the position of having to interact with a computer and allows the attorney to concentrate on communication. In some respects, voice recognition programs today both foster and hinder this objective. The hindrance is as a result of word recognition errors, the importance of correcting error to avoid future inaccurate word recognition, recognition of "discrete speech" rather than "continuous speech" and probably to a lesser extent, simply the speed of dictation. Discrete speech means words for the most part are recognized only if there is a pause between words rather than recognizing complete sentences. Voice recognition programs which allow continuous speech recognition are now in development but are not yet commercially available. Discrete speech voice recognition programs recognize phrases and stock sentences only if specifically trained to do so. On the other hand, even currently available discrete speech voice recognition programs do permit one to sit back in one's chair, simply speak into a microphone and, in effect, talk to the computer screen. One might think that discrete speech might slow down one's train of thought. Surprisingly, speaking with pauses between words does not seem to hinder significantly the continuous thought process. In fact, it almost has a tendency to enhance the thought process with the discipline involved and the resulting tendency to dictate in a monotonous tone.

Unfortunately, there is a psychological stress and anxiety introduced in dictation by a concern that movement may result in the recognition of noises as words. For example, the brushing of a mustache or a beard against the headset microphone will have a tendency to create false words. So may a cough, a sneeze, a rustling of papers. The ringing telephone can introduce additional noise and spoken words before you have a chance to shut off the microphone. The voice recognition program does slow down communication in comparison to using a dictating machine. It is particularly a problem for the lawyer who speaks quickly into a dictating machine. But for the lawyer who is more contemplative or perhaps more considerate of his or her secretary or transcriber, this procedure may not be an unduly inhibiting procedure in creating a document.

It only seemed appropriate to prepare this article by creating a rough draft using DragonDictate. The program being a Windows program integrates well with other Windows programs. For those familiar with Windows programs, installation of DragonDictate creates its own Windows program group and all programs to run with DragonDictate need to be copied as program items into its program group. This allows any Windows program within the group to be called up by speaking voice commands. A voice command is a word or group of words spoken which causes the computer to perform a particular task. For example, by saying "bring up" followed by "Timeslips", the Timeslips program could be brought up on a screen. "Bring up " followed by "WordPerfect" causes WordPerfect to be brought up on the computer screen. DragonDictate permits dictation of text into virtually any Windows program directly. For example, part of this article was dictated directly into Echo Professional, (a personal information management program) while part of the article was dictated into WordPerfect 6.1 for Windows. By using the verbal command "Change Window" the computer brings up the next loaded Windows program as if one had typed the Windows command "Control Escape". As with using a dictating machine, dictation works best, at least for those of us less than genius lawyers, where there is some degree of planning and a few notes are made before commencing dictation. If does not take too long a time to get used to speaking in discrete speech, that is by pausing between words.

What type of computer equipment is required to be able to use DragonDictate? As a practical matter, a fast computer with a considerable amount of memory is needed. I have tried using the program on a Pentium 90 MHZ machine having 16 megabytes of memory and found this to be adequate to run both DragonDictate, WordPerfect for Windows 6.1 and one or two other programs at the same time. Any less memory is not recommended and more may be helpful. The program requires approximately 25 megabytes of hard disk storage for a single user of a computer and somewhat more storage space for additional users on a single computer. In addition, DragonDictate also requires a readily available sound card. You may be aware of the existence of sound cards through your children's game and multi-media computer programs received during the holidays. A sound card is an electronic circuit typically costing one hundred to two hundred dollars. They are sold at local computer stores, electronic stores and even toy stores. The sound card is necessary to convert sound waves from a microphone to digital information that the voice recognition program can understand and convert to words.

Only last year, discrete speech voice recognition programs came only with their own special sound cards. DragonDictate no longer requires its own special sound card to be able to use its program. Rather, it certifies certain sound cards as being workable with the program. The computer used to evaluate this program was purchased with a sound card included. A computer which is sold with a CD-ROM and speakers is virtually certain to also include a sound card. Nevertheless, not all sound cards will work with DragonDictate. For example, when I tried to use the sound card that came with the computer, my reaction to the program was negative since the word recognition rate was very low. After calling technical support at DragonDictate, it was suggested that perhaps a different sound card such as a Sound Blaster 16 should be used. After some effort, I switched sound cards. While the cost of the Sound Blaster sound card was only about a hundred dollars, it does take effort to install. Installing a sound card requires the removal of the computer case. Once the card is physically installed, software that comes with the card must also be installed. If you are on a network or have other cards installed in your computer, it may take additional effort in order for the sound card to work properly without conflicts. After changing sound cards the recognition rate for spoken words increased significantly.

The particular version of DragonDictate which I reviewed is the Classic Edition. That particular version has as a suggested retail price of $695 which includes a headset type microphone and includes a built in dictionary of 30,000 words. The typical spelling program in a word processor has a dictionary of more than 100,000 words. While the number of words seemed much less than in a word processor spelling program, it was rare when a needed word was not in the dictionary of DragonDictate. On contrary the contrary, many rather sophisticated words were recognized as were a number of first names, common and even not so common last names (e.g. Garcia, Gillman, Gould, Gottlieb, Grossman), cities (e.g. Los Angeles, Pasadena, Glendale, San Diego, San Francisco, San Jose) and basic words in the lawyer's vocabulary (e.g. estoppel, defendant and counterclaim). Other versions of the program include a limited version with five thousand words for $395 and a power user version with 60,000 words for $995.

Before actually using the program for dictation it is necessary for the program to recognize the voice of the person who is using the program. This requires a training session which lasts perhaps as little as to 20 minutes. During the training session, four collections of words and word groups are spoken which appear on the screen. The program extrapolates the pronunciation from the trained words to be able to identify other spoken words which are not trained. A separate voice file is created for each person using the program and each such person must go through the 20 minute training. The person using the program must work only with his or her separate voice file.

While DragonDictate is most beneficial for dictation it can also be used to edit documents. The program can facilitate the use of spell checking programs and grammar checking programs. In using DragonDictate for editing, "word right 5" moves the cursor to the right 5 words. "Down" causes movement of the cursor downward. "Shift key end key" highlights a block of text to the end of the line and then saying "delete key" deletes the highlighted test. Some amount of keyboard interaction is usually helpful as well, but the combination of both editing by hand and with DragonDictate does facilitate editing. DragonDictate is very helpful for inserting substantial amounts of text verbally. It takes some effort to get used to moving around on the screen using voice commands rather than with a mouse. Although initially awkward, using voice commands particularly when combined with the use of a mouse and manual entry, can be an efficient way of editing a document.

One characteristic of DragonDictate which has a tendency to slow down the flow of information from the lawyers mind to the computer screen are the errors in word recognition during dictation which do occur frequently. The program is designed for correcting errors as you dictate rather than later. Since DragonDictate continually modifies and updates its recognition of words based on words which are spoken, it is important to make corrections while dictating or else the recognition accuracy will eventually be degraded. Certain noises tend to be recognized as words. Thus, the word "to" appeared on the screen on numerous occasions when no words were spoken.

When dictating, after a brief pause, a choice list appears which lists up to 9 words with the 10th selection being the three question marks (???) If the word appearing on the screen, which is the first of the choice list, is correct the choice list is ignored by the user. The correct word can be selected from this box. Selection is made by saying "choose 6" for example to select the sixth word in the choice list. Recognition errors are not infrequent. Initially it is disconcerting to be continually interrupted by these errors. After using the program for about one week, the interruptions caused by errors both seem to be less of a problem both because the number of corrections tends to be fewer and because corrections can be made quite rapidly. Yet it still slows down the dictation process and of greater concern, interrupts one's train of thought.

Corrections prior to the last recognized word are made in text on screen by saying "oops" after which a row of previously recognized words appears, referred to as the word history. The words which have been input to the screen are further identified by word number, e.g. "Word 1", "Word 2". Underneath "Word 2" is one of the last words recognized, beneath which is a drop down box listing a group of up to nine words, any of which can be selected in place of the word which previously had appeared on the computer screen. Selection of any of the words in the drop down box can be made by saying "choose 1" or "choose 2". After practicing with this procedure for even a week making corrections in this manner does become second nature. In the event that the desired word does not appear in the choice list, the tenth selection showing is three question marks, "(???)" is the selection to make. This means in effect, none of the above, and erases the word recognized from the dictated text. If one of the selection of words seems somewhat close to the word desired, instead of selecting the question marks, one can say, for example "edit 4", and that word will rise to the top of the drop down box. Usually modifying that word with a few letters at most is enough for the program to recognize the word you are seeking.

When the choice list appears and if the correct word does not appear in the list, it is possible to say "spell mode" and then spell the word with what is referred to as the Alpha-Bravo alphabet in which one word is spoken for each letter being spelled, for example, Alpha representing "A", Bravo "B", Charlie "C", Delta "D". A full word must be spoken for each letter to be recognized. For example, to spell the word "defendant", one would say "Delta Echo Foxtrot Echo November" and by this time the words in the selection box will have changed and the word "defendant" is likely to be the first word in the selection box. However, it is only possible to change a limited number of words in the word history prior to the last word dictated. Therefore it becomes important to correct mistakes immediately after they occur. This means that you are forced to keep your eyes on the computer screen rather than solely on your desk, your pleadings or the window or ceiling in your office.

A number of commands which are generic to dictation as well as commands which match specific programs can be spoken in order for those actions to appear on screen. To for example, in order to save a file in WordPerfect for Windows, it is only necessary to say "command mode" followed by "file" and "save". By saying "file print print" WordPerfect commands are actuated and the document is printed from the screen without any further human intervention.

There are several unique words and phrases which are used in the program which might be annoying to those walking by one's office. They conclude "oops", "go to sleep" and "wake up", the latter two phrases being used to turn off and on the microphone and thus the recognition of words. "Scratch that" causes the last word spoken to be erased. The program uses various conventions which, for example, automatically recognizes the creation of beginning and ending of synthesis, capitalizing the first word of a new sentence, and spacing the new sentence from the last period.

DragonDictate allows you to create voice macros in any Windows program which you may be working in. The macros can be useful to insert substantial quantities of text material, such as various boilerplate paragraphs for a contract, frequently used sentences, frequently used names and addresses and various ways of controlling any aspect of the computer. For example, you could say "my address" and have the program type in your entire name and address, or say "declaration form" and have the first and last paragraphs of a form declaration typed automatically. It is difficult to remember all the various commands that might be used for editing and using DragonDictate. To help out, DragonDictate provides a foldout quick reference card listing various voice commands and the "Alpha-bravo" words, that it the alphabet which uses unique words to spell each letter. It is also helpful to have one's own reference card showing the various macro commands that have been created to make it simpler to dictate.

DragonDictate allows the program to be fine tuned to a number of user preferences. The number of words which are kept in the word history which is brought up by saying "oops" can be increased to as many as 32 words. It is also possible to set up DragonDictate so that it adapts to your voice only upon correction of words. That might be useful after the program has been in use for a considerable period of time. The size of the word lists in the drop down boxes can be enlarged so as to be visible on the screen from a considerable distance. The degree of word recognition versus noise recognition can be adjusted, along with a variety of other ways of the program to one's particular needs.

Is DragonDictate a panacea to the practice of law without a secretary? No, but it is a significant and useful lawyers tool. Computer power, software sophistication and reduced cost have now come to the point where voice recognition programs are now practical. DragonDictate is probably not well suited to the lawyer in many large law firms with continuous and substantial support, but despite its limitations probably is ideal for the lawyer in a small form or for sole practitioners who are interested in seeing work product immediately on the screen upon dictation.

Suggested Title:
 
 
 
 
 
 

"Can You Hear Me?"--
Dragon Dictate for Windows Minces Words for Your Office
 
 

By: Paul D. Supnik

Biography:

Paul D. Supnik practices in Beverly Hills in the areas of entertainment, copyright and trademark law. He is a past chair of the Intellectual Property and Entertainment Law Section of the Los Angeles County Bar Association.

_____

Summary:

The article reviews the practicalities of using a voice recognition program, DragonDictate for Windows 1.0, in a law office. The program allows the dictation of words directly onto the computer screen in Windows programs. The practical aspects, advantages and limitations of using the program in a law office environment is discussed. Despite significant limitations, the program has an important value to the lawyer today. DragonDictate 1.0 for Windows is available from Dragon Systems, 320 Nevada St., Newton, MA 02160, 617-965-5200.