According to Peter Mahoney, senior vice president and general manager at Nuance Communications, "The average consumer types about 25 to 30 words per minute." For those people -- and for those for whom carpal tunnel or other disabling factors prevent efficient typing -- being able to talk to your computer sounds like a good idea.
Nuance's Dragon NaturallySpeaking 10 software offers an alternative to the speech recognition feature that comes with Windows Vista and Windows 7. (The company also offers MacSpeech Dictate, which it acquired in February, for Mac OS X users.)
Dragon currently comes in a five editions: Standard ($100), Preferred ($199), Professional ($900), and specialized versions for Legal ($1,200) and Medical ($1,600). The Preferred Wireless edition ($300) includes Plantronics' Calisto Bluetooth headset, while Preferred Mobile ($250) includes a Phillips digital voice recorder.
The various editions are supersets of the Standard package, meaning they all include dictation and voice-control capabilities. The Preferred edition adds shortcuts and the ability to translate from a handheld recorder, while the Professional version adds network administration, the ability to combine commands, and better integration with applications.
For this review, I tested Dragon NaturallySpeaking 10 Professional Version 10.1. But most users could make do with the significantly less expensive Standard or Preferred editions.
All editions of Dragon include a headset. On its Web site, Nuance refers to it as "a high-quality noise-canceling headset," but given that it's included in the $99 Standard edition, I was skeptical about how "high-quality" the headset could be. But it seemed to carry my voice input adequately.
Dragon NaturallySpeaking installs quickly and easily.
Before you begin using the software, you have to calibrate it to help it work with your voice and speaking style. You start with short tasks to confirm good microphone placement and volume settings. Next, you select your accent (such as "Inland Northern U.S.," "British-accented" or "Spanish-accented English"). There are short passages you can read to help the application adjust to your voice, and different vocabulary types you can choose from, such as "General" (U.S. English), "Commands-Only" (for control but not dictating) and "Teens."
Dragon can also improve its recognition accuracy by data-mining your e-mail and contents of your My Documents folder, which helps Dragon build its list of words and phrases you use, including things like the names of people you e-mail.
Wrestling with the Dragon
When you start it up, Dragon presents a small floating toolbar. Once the microphone is active, you can dictate text, control Dragon (including telling it to "go to sleep" and "wake up") and control applications. Dragon can work with several applications, including Mozilla Thunderbird, Lotus Notes and Microsoft Word, Outlook and Excel.
I tried Dragon with samples from a variety of texts, including Green Eggs and Ham, Alice in Wonderland, Pride and Prejudice, An Outline of Science and Computerworld's own About Computerworld page.
As you speak, Dragon displays what it believes you've said in a little pop-up window, modifying the contents if subsequent words make it change its mind about what you've already said. It then drops the text into DragonPad (a customized version of NotePad) or into a field in the current application.
You can also give commands. Dragon has a predefined set of command words and phrases, and you can also create your own. For example, saying "Cap" capitalizes the first letter of the next word, "Open paren" starts a parenthetical clause, and "Scratch that" deletes the previous word. Dragon also has commands to navigate around the screen, select text and more. There are keyboard overrides to force or avoid something being interpreted as a command.
I found it a bit difficult to use (and remember) Dragon's voice-actuated commands -- and to get the hang of Dragon's pacing. I spent around 10 hours with Dragon NaturallySpeaking -- enough to verify that the software does, as claimed, enable you to perform voice entry and editing of text and voice command of Windows and applications. But I'm far from proficient and productive in it. The documentation acknowledges, "We've found that it can take some users about 4-6 weeks of regular use to reach the highest levels of accuracy with Dragon."
Aside from the technical challenges of converting analog sound to text, there's the human aspect. You've got to train yourself to speak in a clear, evenly paced fashion that's oriented toward phrases and sentences (Dragon can be set to autopunctuate). You've got to learn and remember the commands -- and have them at the tip of your tongue. It's easy to become frustrated -- for example, when you make a mistake and blurt out "Oops," instead of saying, "Scratch that." You then have to delete several words -- and "Scratch that" won't work if Dragon doesn't interpret it as a command.
Why? Because timing is important. Dragon is sensitive to micropauses in trying to decide whether what you said was literal text or an attention-getting command. You also need to learn to speak in medium-long phrases, since Dragon's recognition engine is trying to guess words in part by context, e.g., to determine if you meant "to," "too" or "two."
In my own trials, Dragon's recognition of my dictation and commands (or, arguably, my mastery of Dragon) ranged from fair to poor. Many words and sentences came out near-perfect; many were nowhere near what I was saying. Forcing literal text with the Shift key worked sometimes but not all the time; in particular, successfully entering the words "new paragraph" involved practice and luck.
Something to keep in mind: Windows Vista and Windows 7 include speech recognition (and it's available for XP, although you may need to find and install it). You won't get as many features as Dragon NaturallySpeaking offers, but since Nuance doesn't offer a free-trial version (although there is a 30-day money-back guarantee), you may want to start there.
In addition, the company offers a lot of bundled and online resources; if you're committed, you may also want to consider working with one of its partners for help with installation, customization and training.
Dragon NaturallySpeaking 10 does turn speech into text, lets you voice-control the dictating process, and works with other apps, such as Firefox and Windows. But expect a long learning curve. For me, typing is still faster -- but I'm going to keep practicing, and also continue exploring more of the features and options. And, of course, I have the option of typing. People who, for whatever reason, can't type are likely to find the hours needed to properly master Dragon NaturallySpeaking are well worth the effort."