Read the Original Article at http://www.informationweek.com/news/showArticle.jhtml?articleID=199905719
My first experience with voice recognition was in the early 1990s, when I hooked a mic into my Mac and began experimenting with a software program that claimed it would let me control my computer by voice.
Determined to master this powerful new genie, I suffered through several days of minor insults and misinterpretations of my spoken commands. Finally one afternoon after coughing while closing a file draw on my desk, I glanced at the screen and gasped, realizing the program had misconstrued my involuntary hack as a command to delete all the files in the active folder -- my hard drive. Pouncing upon the mouse, I pressed the cancel button in time to avert a minor catastrophe, and then, without skipping a beat, removed the program from the system.
While not nearly as dramatic, my more recent encounters with "speech to text" software have convinced me that this technology wasn't yet ready for prime time. However, after a friend convinced me that he knew someone using Dragon NaturallySpeaking to effectively dictate letters at work, I decided it was once again time to chat with my computer and discover whether I, too, might enjoy the convenience promised by this technology. I obtained a copy of Dragon NaturallySpeaking 9.0, Preferred Edition, and loaded it onto my year and a half old Hewlett-Packard laptop running Windows XP. (There is an upgrade to version 9.5 on the product Web site for those users who are running Vista.)
An Impressive Start
The good news is that, within a half hour of set up and learning the absolute basics on how to work with Dragon, I was able to hook in a mic (one is provided with the package) and have my speech accurately translated into text. I found that most sentences were error-free or at least reflected my intent. While there were still enough mistakes to get an occasional chuckle, it was hard not to be impressed with the software’s accuracy. If you need to do a brain dump to capture some initial thoughts about a new project or to dictate a first draft of a letter, Dragon should perform well right from the start.
Things get tougher if you're hoping to complete your document with minimal use of a mouse or keyboard. At that point, you'll need to invest some real time in learning how to use the software.
Dragon NaturallySpeaking offers suggestions when it gets a phrase wrong.
(Click image to enlarge.)
This is not to minimize Dragon's rich set of tools and commands, and a moderately intuitive design that allows you to select text to correct mistranslations, edit sentences, and make format changes. Most remarkably, if you spend some time correcting translation errors, Dragon will learn how you speak and improve its accuracy as you work with it. For example, if you say "look into mobile recorders," but Dragon records "looking to mobile recorders," all you need to do is say "select looking to mobile recorders" and Dragon will select the text and display several choices of what the corrected text might be. You can choose one of the suggestions or simply restate the words a bit more clearly.
If the program still has trouble you can select the text -- for example, "select Dick Tatian" -- and then say "spell that," which will bring up a dialogue box for you to start inputting the specific letters by voice or keyboard. As you input the letters, Dragon will offer different choices in the spelling dialogue box, so you can cut things short and say "Choose 2" if the correct selection appears.
Once you've corrected a word, Dragon will usually get it right the next time it's spoken. According to the company, this error-correction process will allow Dragon to improve its level of accuracy towards 99%.
However, while Dragon can impressively parse your sentences and determine some words by the context in which they are spoken, it can't understand the context in which you, a multi-faceted, multi-tasking human being with your own individualized work style, operate. To realize the full benefits of the program, you have to adapt to the software and change the way you work.
For example, after an initial infatuation with the program's capabilities, I grew frustrated at my inability to remember the specific voice commands for simple editing tasks that have become second nature with a point and click. Furthermore, I needed to make my formatting requests with a certain rhythm, or Dragon would assume it was part of my input. "Scratch that," the command to erase what you just dictated, "go up 5 lines," or "bold last sentence," might be included in the text if spoken too tentatively. Yes, input was now easy, but editing and navigation still required more familiarity with the tools.
Simply stating "What can I say?" will display a side window with a list of voice commands applicable to the application you are in, but since I required a bit more assistance, I obtained a copy of VoicePower, a third-party help system developed by VoiceTeach, to provide me with an easier access to voice commands and hasten my acclimation to the new environment.
| Dragon NaturallySpeaking 9.0, Preferred Edition
Nuance Communications Inc.
Summary: The leading desktop voice-recognition software allows you to significantly enhance your productivity by incorporating dictation and vocal commands into your computer interactions.
As my comfort level grew, I also became more cognizant of the fact that I didn't need to do everything by voice. Chris Strammiello, director of product management at Nuance, stated that the company recognizes that the program is part of a multi-modal environment. "Dragon is highly accurate right out of the box and the more familiar you become with voice recognition, the less you will need to rely on other input devices," he said. "Still, at this point we don't think desktop users should plan to throw away their mouse and keyboard."
Which Edition Is For You?
There are several levels of Dragon NaturallySpeaking to choose from. The Standard edition, available directly from Nuance at $100, is tightly integrated with word processors such as Microsoft Word and WordPerfect, as well as browsers such as Internet Explorer. Most other popular Windows programs that are based on text input also will accept Dragon-enabled dictation. However, Dragon will not necessarily be able to access all of the menu commands or edit text after it has been dictated.
For $200, you can go up a notch to Dragon NaturallySpeaking Preferred, in which several additional popular applications have "Specialized Application Support" including Excel (though not PowerPoint or Outlook, which require the Professional Edition). In addition to desktop dictation, the Preferred version is able to convert files from handheld digital recorders or PDAs into text, thereby increasing your mobility. Dragon also retains the audio input during an editing session so you can play it back, a helpful feature in cases when you review your document and are somewhat confused by the translation.
If you need to save mobile dictation files between sessions, import specialized vocabularies, build macros to execute custom commands, create a roaming user profile for use on different machines, or utilize other specialized features, you'll need to purchase one of the higher-end products such as the Professional Edition, which has a suggested retail price of $900. Dragon's Legal and Medical editions, which contain specialized medical and legal dictionaries built into the package, run about $300 dollars over the Professional Edition.
There are a number of VARs listed on the Dragon site who sell the software at discounted rates or offer mobile voice recorders, microphones, training, and scripting services to build upon the program. And despite Nuance's assurances that the mic packaged with the software is all you'll need, many of the users who frequent the discussion board maintain that higher-end mics will increase accuracy, particularly in noisy environments.
Is It Time To Speak Up?
For those unable to use their hands due to physical disabilities or limited by extremely poor typing skills, Dragon can be a lifesaver, providing a gateway toward full computer productivity. Nuance offers hundreds of examples on its somewhat choppy Web site of ways in which Dragon is being applied in mobile or chaotic settings (think of an ER situation) where hands are occupied. And with an increased focus on mobility driving Nuance's recent acquisition of several voice technology firms, expect to see a rapidly expanding array of phone-based voice recognition capabilities to input text, purchase products, search for information and access self help services on the growing mobile Web.
Meanwhile, technophiles seeking the brave new world promised by natural computer interfaces will be excited at the level of accuracy provided by Dragon NaturallySpeaking and its utility as a core feature in the nascent immersive network. Writers, bloggers, and even people who regularly respond to large numbers of e-mails also should seriously consider a purchase. In fact, it might just about be time for regular computer users to consider whether being able to chat with their PC might improve their productivity and encourage a more creative flow.