Dragon Naturally Speaking 10: The Review

This time around, the big betterment comes in the quality section. Rather of adding a kajillion new features (cough every variation of Microsoft Word pre-2007) or radically revamping its interface (cough Microsoft Word 2007), Nuance says that lots of its growth effort went into furthering Naturally Speaking’s core recognition engine.

The software now makes 20 pct less recognition errors”though, since version 9 was already amazingly right, a 20 percent boost won’t be almost as noticeable to users as it would have been five yrs ago, when words identification was just coming into its own.

But the identification engine wasn’t just tuned to make fewer slips; it was also tuned to answer more quickly to user input. While edition 9 offers high levels of accuracy, there would oftentimes be a two- or three-second wait while the program decoded your last phrase and spat it onto the screen.

In version 10, Nuance claims to achieve a 50 percent reduction in this kind of response time. Because response time in version 9 was quite an various, looking on decided software package settings, the load on the machine’s CPU, and what was being said at any primary moment, it’s a tough claim to verify. In my own social testing, still, it’s clear that version 10 turns words into text with substantially less delay than its predecessor.

When a platform costs as much as Naturally Speaking 10, many users will object to the lack of big other features”and there’s a case to be made that Nuance’s own innovation engine is sputtering a bit when it comes to NaturallySpeaking (that case is given at the close of the review). But, after 10 versions, this is a feature-rich software package already.

As a steady user of NaturallySpeaking, I’m more than pleased to see an accent on core functionality at the write down of new characteristics. Besides, many of the program’s characteristics relate to correcting errors. If NaturallySpeaking makes fewer errors to begin with, and is more responsive while running, such features are needed less often and thus seem a bit less key.

The recognition engine in edition 10 is zero short of stunning. You can literally set up the software, run through a five-minute developing process, and start dictating with near-total accuracy. If you’ve never finished it before, the action will likely leave you staring in goggle-eyed question at your screen, sputtering sentence after sentence into your headset in an attempt to experience what the platform can’t do.

I’ve been working with NaturallySpeaking for years, so the honeymoon freshness passed long ago, to be exchanged (as in each marriage) with warm fondness and the periodic murderous urge. Why the apoplectic outbursts? Version 9 didn’t support Vista, for one matter, so upgrading to the new OS meant allowing Naturally Speaking down for several months until Nuance put out a major patch. The software just worked with particular variations of Firefox, so after advancing my web browser, I was driven to use Internet Explorer if I wanted to do some hands-free browsing. At Long Last, version 9 brought out my Hulk-rage by corrupting my voice profile, not once but several times”and across dual machines. At Long Last, after gathering hundreds of megabytes of data under a special Nuance program that would “tune” your profile in replace for all that wonderful speech data, Nuance never actually bothered to send me the tuned profile. Amazing.

It’s too early to say if version 10 will fare much better on that last count, but in some ways, it hardly matters. Even in version 9, I quickly realized that the best approach was simply to create a new profile; the software was so accurate that “starting from scratch” wasn’t a big deal. With the improved recognition engine in the new version, it appears to be even less of a worry.