Review by Cthulhu

Dragon NaturallySpeaking’s ability to recognize and properly interpret spoken words within the context they are spoken makes it a useful tool when it comes to dictation (speech to text), controlling a Windows computer, interacting with applications such as Internet Browsers and search engines, the Microsoft Office suite of apps and, through proper customization, just about any Windows app.

After a couple of weeks of experiencing the Dragon, words that come to mind to describe my impressions would be: well-thought, mature, polished, complex, depth, cutting edge, features rich, trainable, expandable, artificially intelligent, versatile. Closer runner-ups would be: essential, time-saver, productivity enhancer, precise, breakthrough - and I may be able to validate these and move them into the first tier of descriptors eventually but a couple more months of daily use may be needed.

I am using the Dragon with my email (Outlook and Gmail), text messengers such as Microsoft’s Communicator, producing quick drafts and with some internet Browser (Chrome) activities such as navigating inside the active tab (page up/down), switching tabs and quick Google and Wikipedia searches. Also some interaction with the Operating System such as opening/closing apps, saving docs and so on. But, of course, there’s a lot more to the Dragon than what I am using it so I will provide what will hopefully be a brief overview based on my own experience.

OPERATIONnnDragon works is by listening to your spoken words and trying to make sense of what it hears. If it decides that the words were a ‘command’, the Dragon decides whether the command concerns the Dragon itself or whether it has to do with the Operating System or the foreground application. Then the command is executed. Otherwise, the Dragon listens to the stream of spoken words and identifies words based on context, statistics of past word usage, custom vocabulary, user-provided training, user profile data, etc., etc. Once the words are identified, the output is produced based on yet another set of rules.

SETTING UP THE DRAGONnnThe initial setup was not exactly ‘quick’ but the no-nonsense thoroughness actually persuaded me that I was dealing with a serious, mature, well-thought piece of software.

One of Setup’s first steps is to identify an ‘audio source’ and Dragon supports most audio inputs a PC supports: USB, Bluetooth, a microphone or a digital recorder plugged into the Line-in jack, sound files, a Pocket PC device used as a digital recorder.

After the audio source is identified, the user will specify his/her age, location (US, NZ, Australia, UK, India, SE Asia…), accent (standard American, Middle America, Southern, Indian, Australian, UK…).

Then, training begins. Volume and the audio equipment quality is tested, then the user is asked to read a specific test. To further enhance its word-guessing capabilities the Dragon then ask for permission to take a look at the user’s text documents (My Docs) and emails.

This completes the initial setup.

TRAINING THE DRAGONnnAs suspected, the setup is just the beginning. To improve word recognition accuracy, the user is expected to constantly fine-tune the application or rather the profile that the Dragon creates for the user.

There are user-initiated tasks such as adding-removing words from Dragon’s written form/spoken form vocabulary listing, by asking the Dragon to take a look at more documents or emails, creating simple new commands that can be universal or application specific or using a full-fledged scripting language to produce more complex outputs from word commands. The user can also train the Dragon on specific words by speaking the word into the microphone and providing the associated spelling.

In addition, the Dragon has the ability to learn and improve its accuracy by periodically reviewing its past activities - there are several tasks that do that periodically while the user-focused portion of the Dragon is not running.

RIDING THE DRAGONnnI confess that I was highly skeptical before trying this out. I expected a low word recognition rate, an unacceptable degradation in my laptop’s performance and apps freezing or crashing while the Dragon was on. I was proven wrong on all of the above.

My very first few sentences spoken into the Dragon were recognized 100% - I suppose the Dragon struggles hard to make a good first impression. Ever since, the Dragon’s ability to recognize my commands as commands and execute them and to properly interpret my words stayed high enough. I am now comfortable to use the Dragon on all my casual emails (Gmail) and on my office emails (I do check before I say ‘send’ and I do enter the ‘to’ and ‘cc’ names manually), on all my instant text messages and on ‘first draft’ office documents.

Again, surprisingly, my computer did slow down appreciably with the Dragon on. This is only my subjective impression because Task Manager tells me that Dragon does take 30-60% of CPU while in listening mode. The good news is that it’s easy to turn the Mic off when Dragon is not needed and then Dragon’s CPU usage drops to 3-4%.

And there is no significant bad interaction with other apps that I’m aware of. I running the Dragon on a Windows 7, dual-CPU Thinkpad T400 and I haven’t detected any ‘new’ problems following the Dragon installation.

THE DRAGON TRAINING/RIDING YOUnnThe Dragon does a lot to make itself useful and helpful. There is an optional sidebar that lists the most common commands available within the current context or you just say ‘what can I say?’ and the available commands show up, ‘Help’ pops up whenever you say ‘give me Help’, a brief tutorial with examples covers most activities, there is a Web site… and so on.

While the above paragraph could have fit under the Dragon Riding section and it’s true that many serious apps provide that kind of help, the Dragon is unique in its training its users into gradually modifying the way their minds work. It’s a gradual process but the fact that the longer the spoken piece is - the Dragon tries to make sense of what you said whenever you ‘pause’ - the more accurate the Dragon is at guessing what your really meant to say. In other words, the Dragon is gently training you into speaking in complete, well-structured sentences.

Speaking for myself, being the fast touch typist that I am, I tend to type first and complete my thought later - because I can quickly erase/rephrase my text. This approach does not work well with the Dragon - it’s possible to correct and modify a text but it’s not efficient. Like I said, I am currently comfortable with using the Dragon for quick emails and messages but not for ‘serious’ documents. I do notice that my Dragon-dictation skills are improving and more of my soundbytes are complete sentences.

MY RATINGnnThe Dragon is one of the few pieces of computer software within this price range that does something useful and it does it intelligently. You can’t have a conversation with the software the way you would with a typing clerk or a computer operator but, as the software evolves (I am reviewing Version 11) and as our little computers get faster and more capable, the Dragon seems to be very much on the path to that level of interactivity.

The way it is now, I will continue to use the Dragon. I don’t expect for everybody to love it or ‘like’ it because it really depends. It does require some significant time investment to make it work well and… read the ‘Dragon Riding You’ section above. I personally like/love this software and, therefore, MY rating is a 5-star but, if I can recommend anything, it would be ‘try it before you buy it’.

–n>> Brush your teeth, it’s the law! <<