Road Warrior

How to Train Your Dragon

Jeffrey Allen

For some time, I have made intermittent use of the wonders of voice recognition (VR) software. Nuance has been the premier provider of VR software for the last several years, ever since it acquired the Dragon VR engine. Nuance now offers VR software under the Dragon label on both the Windows and the Mac platforms. Although the Windows and Mac versions have substantial similarity, more effort has gone into the Windows versions, which has resulted in the Windows versions reflecting a bit more sophistication and refinement. Most importantly, on the Windows side, Nuance released versions designed for use in the legal profession. The primary difference between the regular and “Legal” versions is the inclusion of a more substantial dictionary that includes legal terminology. Everything comes at a price, however, and the Legal version costs significantly more than the regular version. I always wonder about the trade-off of the time to teach the program the legal terms I normally use and those in the legal vocabulary included in the Legal version. Whether you will get good value for the increased cost of the Legal version depends in large part on the overlap between the provided vocabulary and the terms you regularly use. Higher overlap provides greater value. Lower overlap means lesser value.

Nuance recently released new Mac and Windows versions of Dragon. The current version on the Windows side is 15 for the individual Professional and Legal versions and 13 for other versions; on the Mac side, 6 is the current version. (For more details, including pricing, see nuance.com.) GPSolo will review both versions at a later time. For now, suffice it to say that the new versions appear to work better and more easily than their predecessors. Accordingly, this appears an opportune time to discuss the process of setting up and training Dragon VR software to work with your computer.

For as long as I have used Dragon products, the process of training your VR software assumed significant importance. The initial training process required to start up the program has decreased significantly over the last several iterations, and the basic training process has come down to taking a few minutes to read something into the computer and let the VR software analyze your reading, speech patterns, voice inflection, and accent. While you can make the software work off basic training, the reality has always been (and remains) that the more time and effort you put into training your Dragon, the easier you will find working with it and the more accurately it will transcribe your dictation. In the old days, when dictation would go to a human secretary who would transcribe it, the longer the attorney and the secretary worked together, the more easily the secretary found it to understand the attorney and accurately transcribe the dictation. Think of Dragon as your secretary. Think of training the software as you and your secretary learning to work together.

Like all good things, training your Dragon will take some time, but eventually it will work well and efficiently. You will find this true on both the Mac and the Windows platforms. The more you use the Dragon software, the more accurate it will grow, if you take the time to correct the program when it misunderstands your dictation. If you do not go through the process of training the program and correcting it when it makes errors (i.e., telling Dragon it made a mistake, not just manually typing in the correct information), it will not learn, and the problem will repeat itself over and over again. If you take the time to correct the program, it will learn and become more accurate.

Watch your language. Dragon comes with a built-in command structure, and you need to learn it to use your Dragon software most effectively. If you do not use a recognized command, then Dragon will not likely do what you had in mind. For example, if you have grown accustomed to waking up Siri with “Hey, Siri” and try “Hey, Dragon,” you will not wake up your sleeping Dragon. The correct command for the Dragon software is “Wake up.” (FYI, Siri doesn’t respond to “Wake up, Siri,” either. You can say “Wake up, Siri” all you want, but Siri will ignore you.) If you want things to go well, learn the Dragon command structure and use it. The more you use it, the closer it gets to second nature, making your interaction with Dragon far more efficient and effective.

Use the right equipment. You can get Dragon to work with lots of different audio and recording devices. While stray noise can still affect Dragon, it no longer seems to suffer from the hypersensitivity that previously plagued it. As a result, you can pretty effectively use built-in microphones on many computers with Dragon. You can also use stand-alone recording devices and upload the audio files to your computer for processing by Dragon. Stand-alone recorders often have diminutive sizes, and you can easily carry them in a pocket, making it easier and more convenient to dictate into them. While you can certainly get away with a lesser recording device, I have always preferred to use professional dictating equipment as I like the control structure and find it easy and comfortable to use. I can’t tell you whether this results from the inherently superior design of those controls or the fact that I practiced law for many years using dictating equipment with that control structure and I simply grew comfortable with it.

A good professional dictation device will cost you about $400 to $500. The two portable devices I like the best come from Philips and Olympus. If you use a Mac, you will want the Olympus; if you use Windows, either one works fine. The DPM8000 ($499, dictation.philips.com) tops the list of Philips’ offerings. The DS-7000 ($499.99, olympusamericaprodictation.com) heads up the Olympus line. You can find other good-quality devices from Philips, Olympus, Sony (sony.com), Grundig (grundig-gbs.com), and others. The primary thing that distinguishes a professional dictation device from other devices that might work is the inclusion of a thumb slide control for record, pause, and rewind controls. Nuance publishes a list of recording/dictation devices compatible with Dragon. Some devices not on the list actually work, but if you plan to go out and get a new device, play it safe and pick one on the list of Dragon-approved recorders.

Make multiple profiles. Dragon will set up a profile for you when you first start to use it. It will continue to add information to that profile as you use it to make it work more accurately. While you can sometimes get away with making one profile work for more than one computer or more than one input device, I don’t recommend you try that. You will probably find that Dragon works better if you set up and build a separate profile for yourself on each computer you will use with Dragon. It will probably also work better if you set up a separate profile based on the input you plan to use on each computer—for example, one profile using the internal microphone on your laptop and a second profile using your dictation recorder. The sound will differ somewhat owing to the hardware of each computer, and you want each profile to give you the most accurate possible transcription.

Quiet on the set! If you want to, you can go to a football game and sit in the stands and dictate, or dictate on a Saturday morning with the TV blaring, the kids screaming, and the dogs barking. But, if you do, you will likely encounter a high number of errors to correct. You will also run the risk of contaminating your profile so that you will not get a high-quality result the next time you use Dragon in a quieter environment. You do not need to construct a soundproof room or Maxwell Smart’s Cone of Silence (for those of you old enough to remember the TV show Get Smart) to ensure that you maximize accuracy, but the quieter the environment you use for dictation, the higher the level of accuracy you should expect in transcription.

Protect your profiles. Remember that all your Dragon profiles have dynamic properties—they continue to grow and learn. Each time you use a profile, the circumstances and sounds in play become a part of it. If you have a bad cold and use your profile while you sound stuffy or nasal or both, you risk damaging the profile’s accuracy. You probably don’t want to do much dictation when you are sick anyway, but should you feel compelled to do so, either set up a new profile from scratch or duplicate your existing profile and rename it before using it so that you can keep the healthy-voice profile intact. When you recover from your cold, switch back to your normal profile. Speaking of duplication of your profile, you will want to back up your Dragon profiles regularly to protect them from loss or damage.

Spoiler alert. We plan on doing a more detailed review of the current iterations of Dragon for Windows and the Mac in the near future. That review will address the improvements and features of the software and compare it to previous versions in terms of improvements. If you have not guessed from this Road Warrior column, I like the product, and the review will be positive.

Jeffrey Allen is the principal in the law firm of Graves & Allen in Oakland, California. A frequent speaker on technology topics, he is Editor-in-Chief of GPSolo magazine and GPSolo eReport and a member of the Board of Editors of Experience magazine.