Saying Makes It So
A new software program turns speech into text without your having to lay a finger on the keyboard.
I am "writing" this column only in a metaphor cool sense. Scratch that. In a metaphor a coal sense. Scratch that. In a metaphorical sense.
Welcome to Dragon NaturallySpeaking from Dragon Systems, the acknowledged leader in a new generation of vastly improved speech-recognition programs, and one of the neatest—though not quite flawless—tools for interacting with a computer since the invention of the mouse.
With a headset-mounted microphone poised at the corner of my mouth, my feet up on the desk, and my hands knotted comfortably behind my head, I am speaking. Not in some obscure geeky lingo. Not. . . with. . . the. . . exaggerated. . . pauses required by previous generations of dictation software. But in easy, conversational English.
But now the best part: The words I utter are being typed, as if by an invisible transcriber, on my computer screen, punctuation included. When I want a comma, I say "comma," and in it goes. The same is true for everything from "ampersand" to "underscore." And not just text, but all types of editing and formatting commands. "Move back seventeen words. Caps on," I say. "Select previous paragraph. Bold. New paragraph." When I want to center a line, I say, "Select line. Center," and it is immediately moved to the middle of the page.
When either the computer or I make a mistake, I tell it "scratch that," and the error disappears. I can even tell the software to perform commands such as switching to previous windows. This completely "hands-off" approach is the main reason why Dragon NaturallySpeaking represents a step above its main competitor, IBM's ViaVoice, which requires that application functions be handled manually. (At press time IBM reported that a new version, ViaVoice Gold, would soon correct this discrepancy.)
Dragon NaturallySpeaking isn't perfect but it is a dogged learner. The Personal Edition has a preloaded active vocabulary list of 30,000 words. If one of the words you say isn't in the memory—"metaphorical," for example—you can either spell it out letter by letter or, by means of a fairly simple process, add it to the active list. There is also a 200,000-word backup dictionary—the same size as a standard, mid-size print dictionary—from which vocabulary can be added. To further acclimate itself to your diction, pronunciation, and jargon, the program's Vocabulary Builder will analyze any document already typed into your computer and automatically add new words it discovers to its active vocabulary. It can even be taught your own idiosyncratic dictation shorthand: If you say "CNN," for instance, it can learn to write out "Cable News Network."
Every time you correct one of its mistakes, the program notes the tweak. After just a few days of use, my miscues dwindled from one or two per sentence to one or two per paragraph. For example, when I dictated the last two sentences, it made two mistakes: "the tweak" was misinterpreted as "to tweak" and "my miscues" as "by miscues." In fact, because Dragon NaturallySpeaking doesn't make spelling errors per se, there's no built-in spell checker, so dictated documents should be carefully proofread.
Setting up the program is relatively painless: The biggest chore is reading one half-hour's worth of material aloud to the computer so that the program can gauge your pronunciation. You can retrain it at any time—and you may have to, for instance, if you move to an office with more background noise, or if your diction is altered by a bout of the flu.
There are, however, two main drawbacks. First, although the Personal Edition of Dragon NaturallySpeaking functions as a simple word processor, it can't be used to dictate directly into other word-processing programs or applications, such as e-mail software. Instead, you must cut and paste into them from the Dragon NaturallySpeaking window. Second, it can only be tailored to a single voice.
But the new Deluxe Edition (which was not available for testing at press time and will list for a hefty $695 as opposed to $150 for the Personal Edition) will solve both problems and include a few extra attractions. It can be used with Microsoft Word, accommodate up to 15 users, and even accept British English, French, German, Spanish, or Italian. It can reverse the voice-to-text process and read your words back to you. Furthermore, it will add more vocabulary while permitting you to increase the active vocabulary (the extent depends on the size of your computer's memory).
Be advised that Dragon NaturallySpeaking, which can take dictation up to 160 words per minute, requires a minimum 133 MHz Pentium machine and 32 megabytes of random access memory if you're using Windows 95, 48 megabytes for Windows NT. (You can't use the program with Windows 3.x and no Macintosh version is planned.) And the more memory you have, the better it will run. Before you buy, make sure to call Dragon Systems or check out its Web site, to ensure that the program will work well with your computer's sound card. The list of compatible cards is long and growing, but if yours isn't one of them the program won't allow you to dictate as rapidly.
Overall, Dragon NaturallySpeaking is just what the doctor ordered for ham-fisted typists, secretary-dependent executives, carpal-tunnel-syndrome sufferers—and just about anyone else who thinks it school to talk instead of type. Scratch that. It's cool.
Dragon NaturallySpeaking comes with a headset microphone and is available at many computer-software dealers. For additional information, contact Dragon Systems at 800-437-2466, or visit one of these sites on the Internet: www.naturalspeech.com or www.dragonsys.com.