Nuance’s Dragon, Version 9: Cloud nine for anyone needing dependable speech recognition software.

December 28, 2006 | By | 11 Replies More

Dragon, for those who are not seen it in action, is a speech recognition software. You talk into a microphone connected to your computer and the program transcribes your words into written text.  For those of you who haven’t seen the process of speech recognition before (or those who have only seen earlier versions of voice recognition software), the current version works like magic. 

Who could make good use of speech recognition software?  Anyone who writes.  That probably includes you.  All of you bloggers, take note.  Same for all of you professionals.  For instance, I work as a lawyer and I spend an hour or two each day using Dragon. 

I have worked with Dragon ever since Version 6. It is now up to Version 9.  I use the “Preferred” version of Dragon 9, which costs about $160.  The software comes with a microphone in the box.  The software is loaded with a vast vocabulary of legal and medical terms, something to keep in mind for those of you who might otherwise be tempted to jump to the vastly more expensive Legal or Medical versions of Dragon. For most people, there’s no need to make that jump, in my opinion.

There are millions of people out there who could make good use of voice recognition software. Dragon allows you to give your hands a break, even if you are a proficient typist.  Dragon works so well that it seems like magic. No, I do not own any stock in Nuance (the company that sells Dragon Naturally Speaking). I’m simply an enthusiastic user trying to share the joy.  BTW, here’s Nuance’s information on Dragon 9 for some basic information on Dragon 9. 

The software packaging indicates that Version 9 is faster and more accurate than typing. This is an understatement, in my experience. According to tests described on the package, most people can dictate 140-160 words per minute. A 900-word document can thus be produced in six minutes. If that same document were typed at 40 words per minute, it would take 22 minutes to produce. In my experience, I suspect that Dragon allows me to cut my writing time in half when I am summarizing meetings and depositions.  There is no comparison between versions 6 and 9 in terms of accuracy.  If you haven’t tried Dragon since Version 6, take a serious look at Version 9.

I use Dragon to produce the initial drafts of most of the long documents I write. I find first drafts to be especially frustrating to key-in. When doing first drafts, one’s mind typically races ahead and sometimes morphs in midstream, leaving one’s semi–nimble fingers behind. It is truly a joy to turn one’s thoughts or scribbled notes into a highly accurate document painlessly filled with text. Dragon is fully capable recognizing one’s voice to make edits too, but I typically switch over to my fingers on the keyboard for the numerous tedious edits that I use to turn first drafts into second or third drafts.

I create the first drafts of almost all of my blog posts using Dragon. It is easy on the fingers. Dragon keeps up with you. It allows you to think instead of focusing hard on pecking away at the keys. There is no need to slow down in order for Dragon to work. As the manufacturer indicates, Dragon is at least as accurate when you read in fast fluid fashion as when you slow down and read in a staccato manner.

I have not done a statistical analysis of the accuracy of Dragon, but based upon my own experience, I can as state that Version 9 is more accurate than anything preceding it. It is not at all unusual to dictate entire paragraphs without any errors at all.

It doesn’t take a lot of training to get the Dragon up to a high level of accuracy. Out of the box, Dragon invites the user to dictate for about 15 minutes as the initial training. At that point, you’re up and running. You can fine tune and tweak Dragon further, of course, and this is highly recommended. For instance, I often spend a few seconds prior to dictating telling Dragon the correct spellings of any new proper nouns that I will be using (the names of people and places). In the course of dictating, Dragon will sometimes not understand a particular word.  At that point I “train” the software by selecting the word and indicating how it should’ve been recognized. These corrections are easy to do. They become second nature to experienced Dragon users.

Here’s a couple of caveats: the manufacturer claims that Dragon 9 works with most commonly used desktop applications such as Microsoft Word or Outlook. In my experience, although the product does function within these applications, it works more accurately when used in tandem with the rudimentary word processor provided with the product, Dragon Pad. I rip through my first drafts using Dragon Pad. Then, I select all and copy that first draft into Microsoft Word for further editing, as well as the spell-checking and grammar-checking that Dragon Pad lacks.

Here’s another caveat: when you write with Dragon, you need to be careful to proofread your documents with the understanding that your errors might be dramatically different than the kinds of errors that you experience when you key-in your text using your fingers. When you key in your work, you might discover an error such as “the” spelled “hte.” When you use Dragon, you’ll see different sorts of errors.  The program is capable of misunderstanding a word, the result being a perfectly spelled but dramatically different (but similar sounding) word than the one you intended. This problem can be solved by making sure that you proofread carefully, something you should be doing anyway.

Dragon has many uses beyond the ones I’ve described so far. For instance, you can use Dragon to control your web browsing, using your voice instead of your mouse.  Dragon is also able to interface with digital recorders. It is therefore possible to dictate your thoughts away from your computer, then to download that information into Dragon for a surprisingly accurate transcription. I’ve tried this a few times and I’ve been impressed.

I am not able to tell you anything about customer service at Nuance. This is because I’ve never used it. I have not had any need to do so (I have read others’ accounts indicating that customer service at Nuance leaves something to be is desired).

I bought a couple of boxes of Dragon for friends as gifts in the past year.  They’ve both appreciated the chance of lessening the punishment they impose on their fingers every day.  I’ve seen several co-workers easily and happily incorporate Dragon into their work lives. All it took was a little encouragement from me.

If you do a lot of writing, consider whether this product will pay off for you. If you dare to give it a try, you might be so incredibly pleased that you’ll be out there preaching the virtues of this product, just as I am doing. 

And, yes.  I dictated this post using Dragon 9.  In the course of dictating my first draft, Dragon 9 made only four mistakes.

Share

Tags:

Category: Technology

About the Author ()

Erich Vieth is an attorney focusing on consumer law litigation and appellate practice. He is also a working musician and a writer, having founded Dangerous Intersection in 2006. Erich lives in the Shaw Neighborhood of St. Louis, Missouri, where he lives half-time with his two extraordinary daughters.

Comments (11)

Trackback URL | Comments RSS Feed

  1. Deb says:

    I am one of the grateful recipients of Erich's largesse (thanks, again Erich), and have been using the software for a couple of weeks now. I was reluctant to try due to an attempt several years ago (version 2?), but very glad I tried again.

    For those of you parents, it's a lot like potty training an 18 month old baby: it isn't the baby that gets trained so much as it is the parent. Once you train yourself to use certain command words, and dictate quickly, it goes much smoother. You can add your own commands in addition to the existing ones- for example, I can now say "signature block" and my entire name, address and phone numbers are automatically entered. At first I was more frustrated than productive, but by now, it is the other way around.

    I'm also learning to use it to move around in different programs/windows, but that isn't nearly so useful as wordy drafting of documents, because the commands are so slow. The program waits to see if you are dictating a sentence and just using the word "select" or "tab" or "click", etc., as part of the sentence before executing the word as a command. Dictating long passages with few commands is very fast. When my hands are really hurting from overuse of the mouse, I use it for the commands even though it is slower, and just use the mouse to add a few quick window changes. Its a great solution, and although I am still not getting as much done as when I typed or dictated to a transcriptionist, every day is a little faster and the avoidance of pain is very much worth it.

    Course now I have a headset for the voice recognition software, with a cord to my left, and one for my phone for the other ear with a cord to my right. When I wear both, I really am tied to my desk!

  2. grumpypilgrim says:

    How well does the product work in the presence of background noise — say, in an airport or noisy office?

  3. Dan Klarmann says:

    I've been following the development of speech-to-text software in the computer tech journals since the early 1980's, and freely admit that I've never bothered to try it. When I'm composing text, my typing speed is generally adequate. In fact, I remember that when I switched from handwriting papers to word-processing (1981), I found that I had to force myself to slow down to keep my composed prose from sounding like the blather that one sees so frequently in the blogosphere. But then, I'm more of a math-brain than a verbal one.

    I'm reminded of an episode from the life of Isaac Asimov. For a while he tried dictating his stories to tape. His typing speed was phenomenal, but he was also a fast talker. The problem was that his excitement in telling his unfolding stories occasionally rendered portions of the tapes unintelligible. He gave it up.

    I also have trouble believing that one can talk a cursor around a browser window as easily as mousing, or even using the keyboard.

    Also, most of what I compose is computer code, where a comma or semi-colon in the wrong place can cause complete corruption, and where few normal words are found.

    That all said, I've only heard good things about Dragon since it was first introduced. Those folks came up with algorithms that may only be surpassed once we leave the realm of Turing programming (rigid rules) and allow heuristic programs (those that learn and edit their own code) on distributed processors to do the job. Programs that evolve themselves exist, but are not yet predictable enough to let loose on the open market.

    I've also always had a tolerate-hate relationship with those little "friendly" utilities that automatically "fix" your errors. One has to watch them like a hawk. When you type, you know what you are getting and notice changes. When you dictate and read it later, you might spend a while trying to figure out what you meant by what Dragon (in this case) thought you apparently said. Unless, of course, you are a trained verbalist, as are barristers (the subcategory of lawyers who argue cases) or salespeople.

  4. Erich Vieth says:

    Grumpy: I haven't really pushed Dragon hard in a noisy environment. I know that using a noise filtering microphone (I use the XVI made by Parrott) filters out most noise. You won't be affected by soft background music or people talking in a distance. I've dictated in hotel lobbies without a problem.

    To put this post in perspective, I should note that I type about 60 wpm. Therefore, I can get the job done with my fingers. I really do enjoy getting first draft text on the page using Dragon, though.

  5. Gordon Good says:

    What microphone do users recommend?

  6. Erich Vieth says:

    Gordon: I know some people who are quite happy with the microphone that Dragon provides with its voice recognition software. I bought a Parrott brand USB microphone. It works well, but I'm wondering whether it is really an improvement over the microphone Dragon provides. I know that out in cyberspace you'll find lots of debate about what microphone works best.

  7. Dan Klarmann says:

    This is how I’ve always expected voice recognition to work in my world.
    I hurt myself laughing:


  8. Erich Vieth says:

    Dan: I couldn’t stand to watch more than four minutes. That was intensely painful. The user was making as many mistakes as the software too. I’ve had my “moments” with Dragon, but never 1/10th as bad as that. It makes me wonder whether the user had properly set up the program and whether he had a good quality microphone. Microphones vary tremendously in quality.

  9. Dan Klarmann says:

    A video reply by a Dragon expert demonstrates how much better Dragon is. It only took him about five times as long to speech-enter that code it as anyone could type it.

  10. Craig says:

    Erich,

    I just upgraded from version 8 to 9, actually 9.5 since I'm on Vista and I'm pleasantly surprised at the improvements. Version 8 wasn't that bad either.

    Regarding the comments about microphones…so far I've tried four different headsets out and the Parrott XVI surpasses them all by a mile. Accuracy is way up.

    Since I'm dealing with RSI, this product has been a blessing.

    — Craig

Leave a Reply