‘Computer, Can You Hear Me Now?’
One Lawyer’s Surprisingly Positive Experience with Speech Recognition Software
By Jim Calloway, Director, OBA Management Assistance Program

In the movie “Star Trek IV, The Voyage Home,” the crew of the Enterprise journeys to Earth’s past (San Francisco, 1986) to search for humpback whales. In one humorous scene, Scotty needs to use a computer. He stands in front of the computer and utters the command, “Computer.” A young man standing nearby sheepishly points to the mouse. Scotty nods his head, picks up the mouse and speaks into the mouse, “Computer,” again with no result.

Computers that operate by voice command have long been a staple of science fiction. Whether this referred to the computers on the starship enterprise, robots or HAL from the movie “2001: A Space Odyssey,” many of us have dreamed about the time when we could utilize our computers simply by speaking to them instead of dealing with a keyboard and mouse.

Well, for many lawyers, that time may now be here, at least in some ways.

This article represents the first article that I have prepared for the Oklahoma Bar Journal by using a headset and speech recognition software rather than a keyboard and mouse. For several weeks now, I have been using speech recognition software to draft e-mails, do postings on OBA-NET, prepare documents in both Word and WordPerfect and execute simple commands.

(And, for those of you who were wondering, there were three mistakes in the preceding paragraph. The program heard “executes” instead of “execute” and failed to capitalize the word “bar” after Oklahoma and the second half of OBA-NET. Frankly, I am surprised that it can be trained to recognize OBA-NET at all.)

I am using a software package that was released in July 2006. It is called Dragon Dictate Naturally Speaking Preferred Edition Version 9. This product has been getting rave reviews and has already been featured both in the New York Times and on National Public Radio. I would hate for any readers to take this month’s column as a product endorsement. But the fact of the matter is that this product is extraordinary and is of potential use to a large segment of the legal community. The software can be purchased from www.nuance.com. Nuance was previously known as ScanSoft.

Speech recognition (or voice recognition) software has been around for quite some time, as evidenced by the fact that the package I am using is version 9 of the product. This is not the only voice recognition package available. There are others, such as IBM’s ViaVoice. It is, however, my belief that at this juncture, one of the various “flavors” of this product is the only product for lawyers to consider.

The version of the software that I am using, the Preferred Edition, retails for $199. In my opinion, the basic version of the product should not be considered for purchase in a law firm. There are two higher priced versions of this product; the Professional Edition retails for approximately $800 and the Legal Edition for $1,000. I’ll cover more about those versions a bit later.

There is no doubt that, for me, this product is a “keeper.” As one individual has aptly noted on the OBA-NET, if you can type 100 words a minute, this product is probably not useful for you. (Although, in fairness, I should note that most people can dictate at a far faster rate than 100 words per minute.) But now that I have been using this product for several weeks, I feel that I would be remiss in my duties if I did not bring it to your attention.

I knew that I needed to give speech recognition software another try some time ago when I saw Oklahoma City attorney Denning Crowe demonstrate DragonDictate Version 8 at a retreat for the OBA Law Office Management and Technology Section. Some of you may also remember that we did a demonstration of speech recognition at our “law office of the future” exhibit at an OBA Annual Meeting a few years ago. It was amazing to me then, and still is today, that speech recognition software could really work at all. After all we have different speaking styles and accents, and there are many homonyms and synonyms which would seem to be confusing to a computer software program.

Denning Crowe also speaks highly of www.knowbrainer.com, a vendor that markets Dragon Dictate and products related to speech recognition and hosts a forum about the topic. They sell advanced command recognition software and legal vocabulary add-ons. I have not purchased anything from them.

Simply put, the difference between this product and all previous versions of it, and other speech-recognition software that I have tried is that this works well enough to be used on a daily basis rather than simply to be demonstrated as a mere curiosity.

So I would encourage the non-typing and slow-typing attorneys who are using their computers on a regular basis to think about giving this product a try. However, there are numerous “catches,” as one might imagine. So let’s take a few minutes to cover some of the ins and outs of why this product works successfully for me.

1) Training – The software developers for this product say that one of the improvements of the newly released Version 9 is that it doesn’t require any training, but will work well right out of the box. This may be true, but my belief is that spending 20 or 30 minutes reading the software voice samples was an exercise that would be repaid quickly by less time spent making corrections. So I began by dictating/reading some of the sample material provided by the software.

2) Your “My Documents folder” – If your practice is not to keep documents that you prepare in the My Documents folder on your PC, you will want to copy many of the documents and correspondence you have drafted over to the My Documents folder before you install the software. The reason is that when the software is being installed, it asks your permission to review all of the documents in the My Documents folder and e-mail applications to familiarize itself with the types of words that you frequently use. I think this is invaluable. When I used the word blogs with the software for the first time, it got it correct because it had noted that word being used many times in my documents. After the initial setup, you can then delete the documents from the folder.

3) Memory – You may need to upgrade your computer’s memory. Speech recognition software is a memory hog. The product specifications indicate a minimum memory requirement of 500 MB (one half of a gigabyte.) Most will find that this is simply not adequate for the task. Less memory causes the display of words to lag further behind your speech, which many will find distracting. This also delays your ability to make corrections. In anticipation of installing speech recognition software, we upgraded my computer’s memory from 500 MB to 1.25 GB. When I have a dozen or so other things open, this still seems to be a bit slow. One of my colleagues who has been using speech recognition software for some time recommends two GB of memory, and I believe that those of you who want to use this product should consider this when you purchase your next computer or upgrade the memory in your current computer.

4) A USB microphone – Those who have made this journey before me are uniform in their advice that a USB microphone should be purchased rather than one that plugs into the soundcard. At the moment, I am using a noise canceling speech headset that I purchased from an Oklahoma City Staples for $29.95. This has worked amazingly well, but I likely will be purchasing a more advanced microphone that will allow me to toggle between the computer and the telephone without changing headsets. I also think a higher quality mic will improve accuracy. In my opinion, a headset is a necessity. While others may want to use microphones on pedestals on their desks, I am interested in reducing as many variables as possible. Therefore I don’t want a desktop microphone because the distance between the microphone and my mouth would vary greatly. That is not the case with a headset.

5) Be patient. I was able to produce usable documents the very first day that I used this software package. However I had some experience with prior versions of voice recognition software. Do not try this for the first time on a brief that must be filed that day. You will not be happy with the results if you start under pressure. I still tend to get a bit impatient and use the mouse or keyboard to make minor corrections. However, it is critical that one uses the software to correct misunderstood words. This is done by selecting the misunderstood word with a voice command and either choosing from a menu of the five closest alternatives or spelling out the word so that the software recognizes it. This is critical because the software learns about your voice and speaking style and improves its database based on your corrections.

6) Proof your documents very carefully. I’ve already had a couple of very minor mistakes slip out in e-mails, even though I proofed them fairly well. A lot of proofreading is looking for misspelled or mistyped words. DragonDictate will not misspell or mistype words. It will, however, from time to time insert an incorrect word. Therefore it is recommended that you slowly and carefully proofread your document so that you catch any incorrect words.

7) Use speech recognition appropriately at first. Think of appropriate ways to use the software while you are learning how to use it. A good use might be drafting memos to your staff or the file where you just want to record your thoughts. It is likely that you will be the only one reading these notes and a word recognition error might not be significant. I have demonstrated the software to a couple of groups of lawyers and several have made the comment to me how beneficial it would be to quickly dictate notes after a court hearing or telephone negotiation. You would still want to proof these, but it would not be as critical in most circumstances if there was a wrong word contained in the document as you would likely be able to understand the correct wording from the context. You might be cautious about using this software to draft a multimillion-dollar contract until you have more experience with it.

8) No, don’t do that! Watch for the command recognition feature to generate unexpected responses. It is really nice that this product allows you to select a word to be corrected by voice command. You can also do many simple things such as sending an e-mail or saving a file by voice command. I often may want to paste text into a document that I am drafting and I really appreciate that this software has a “paste that” command to paste from the clipboard. I have, however, noted the possibility that dictation might be misunderstood as a voice command with negative results.

For example you can send an e-mail by using the command “send.” (A command is generally distinguished from normal voice dictation by a brief pause before and after the command.) However, it is certainly possible that the combination of an inadvertent pause while dictating a phrase using the word “send” or the computer misunderstanding another word as “send” could cause the e-mail to be sent prematurely. I try to guard against this by not completing the “To:” line until I’m finished with the e-mail and have proofed it. When I am doing a reply, I will cut the address out of the “To:” line and keep it on my clipboard (To do that, highlight it, then use key combination Ctrl plus X.) When I complete the e-mail I click my mouse to put the cursor in the “To:” line and use the voice command “paste that” to insert the address. Then I use the voice command to send the e-mail.

If the software incorrectly indents or otherwise reformats a document, I have found that it is easier just to use the undo key command (Ctrl plus Z) than to try to get the software to correct those kinds of mistakes by verbal command.

The product does have a “go to sleep” command feature to temporarily disable the microphone when you are interrupted or the phone rings.

9) Don’t forget to save. It seems to be easy to forget to save your document frequently when using voice dictation. I dictated the first two pages of this article before I saved it for the first time. Use that Ctrl plus S keystroke command to save early and often.

I could go on and on with my recent experiences with this product. But I have noted that one unfortunate byproduct of speech recognition software is the tendency to dictate longer documents and e-mails. I would caution you that people expect e-mails generally to be brief.

One important note: I would be remiss in my duties if I did not note that the new Windows operating system Vista, which is now slated to be released in January 2007, comes with built-in speech recognition software. Industry insiders tell me that this product will be really good for a first release, although not as good as Dragon Preferred Version 9.  I know that some of you will want to wait for a possibly free version, but I am certainly happy that I am realizing productivity gains right now and would encourage others to do so. But this is another factor to consider if you plan on buying a new computer in the first quarter of 2007.

When I did my first blog post about Dragon Dictate to Jim Calloway’s Law Practice Tips, I made the comment that I was not sure that the more expensive Legal and Professional editions would be worth the extra cost. After further examination and a lengthy discussion with the software vendor, I am rethinking that position. Obviously there is a huge amount of difference between buying $1,000 software package and a $199 software package. So I think most readers would be well served to purchase the cheaper version and see how it works for them before making the much greater investment. However, there are some really nice features with the advanced editions. If you decide to use this product daily, then the investment for the more expensive version might well be in your best interest.

For example, with the Preferred Edition, I’m the only individual who can proof the work. This is because when it inserts the wrong word into a document, I am generally the only one who can determine what I meant to say. The more expensive Professional Edition will allow the voice dictation file for the document to be replayed while someone reads the document. This allows a third party to proof the document that the lawyer dictated. There are several other enhanced features. The Legal Edition has those features as well as a built-in legal dictionary and the ability to interpret legal case citations correctly.

I also might note that some people are simply better at dictation than others. If you’re dictation style is to say “the burden of proof on the entity… I mean agency…. I mean party” then you will get just that text whereas the trusty human transcriber will type what you meant to say. After a while, you will be able to dictate quite quickly, but you still must pronounce and enunciate well.

One additional advantage that I did not anticipate is that voice dictation is less tiring than typing all day. I still tend to use the keyboard for final edits and a good number of short e-mails. But when I am working on a lengthy CLE paper, the physical difference is noticeable. I’ve only had the product a few weeks, so I have still not tried some of the other advanced features such as voice dictation into my Treo.

Some of you will find this new software package to be a great addition to your practice. I suspect that not everyone will have success with this product, but hopefully by following the guidelines noted above, you too can be dictating successfully into your computer.

Now, if someone could only build that transporter from Star Trek.

Originally published in the Oklahoma Bar Journal Sept. 2, 2006 - Vol. 77; No.24