I previously wrote a (non) review of Dragon Dictate for Mac 2.0 discussing the history of voice recognition and why I probably would not be getting the newest version of Dragon Dictate.

I have been a fan of speech recognition since just before natural language models. I remember when discrete language ruled the day and you, had, to, talk, like, this, to have any chance of the computer being able to understand your speech. In the last few years, I got tired of paying $150 every 12 to 18 months to try a new version, just to be disapponted with it. The recognition was quite good, but the overall package (how easy it was to correct and train words) dragged it down. I dropped out of the game of regular updates and decided not to update for a while. Well, after a glowing review from Ernie the Attorney, I decided to give it a shot.

I have been using Dragon Dictate 2.0 for Mac for the last few weeks. I have dictated a several articles using it and I have been quite impressed with where Nuance and MacSpeech have made significant improvements. However, some problems remain. [Read more...]

CC photo credit: debaird™

I’m telling you from the start that this is not a review of Dragon Dictate 2.0 for Mac. Since I am including a disclaimer, there really should be no complaint that no review follows. :)

So, it’s LINK BAIT!

I prefer to call it a topical musing on the history of voice recognition and my hopes for the new Dragon Mac offering, and leave it at that.

CC photo credit: secretlondon123

I have been using voice recognition products since 2000; first on the PC and more recently on Apple computers. Back in the olden days, there was a real issue with (sub-gigahertz, single core) processor speeds. Now, my cell phone has more power and memory that my 2000 era computer; and can do a better job with voice recognition. To be fair, today’s cell phones use a client / server model for voice recognition. The iPhone or iPad does not do the voice recognition itself. The Dragon Dictate app beams the voice file to a server farm to churn out what it thinks you said and beam it back to your device.  It’s pretty amazing that our networks are fast enough for this to be practical. I still remember visiting the computer lab at my brother’s grad school where you had to plug the handset of a telephone into a suction cup. That was the era of Net Trek and MUDs, and beer bashes and dancing to “Rock Lobster.” Yeah, um, don’t tell my parents.

Back to voice recognition: over the years, I have bought the new version of the voice recognition software du jour, try it for about a month and then leaving the scene for a year or two. Then I am lured back with the siren call of improved accuracy, speed and functionality, and the promise of finally being able to quickly and naturally enter text into the computer proves irresistible and I open my wallet again. And so it has been through Dragon Naturally Speaking, Nuance Naturally Speaking, MacSpeach Dictate, and now Dragon Dictate 2.0 for Mac.

Then as now venders have touted:

• Accuracy over 97%. The current version of Dragon Dictate 2.0 for Mac touts a 99% accuracy out of the box. Frankly I don’t see the improvement over the years. It’s not that voice recognition is not that good. It is just that it has been around 97% for a number of years now. No one was touting 75% accuracy back in 2000. Even then, venders advertised better than 95% accuracy and it was well in the realm of possibility to achieve it. Certainly, the amount of training required has decreased. I recall reading excerpts from Arthur C Clark’s 3001, or Dave Berry in Cyberspace, that lasted well past the point of any entertainment value. Yes, there is an enormous value to getting a high degree of accuracy right from the start, but engine accuracy has been good for a number of years now. There is still a high degree of training involved: users must stop slurring words, and the program needs to learn the names of organizations so that “Peak Vista Community Health Center” does not come out as, “peak vista Community Health Center.”
• Improved speed. From a getting-words-to-the-page perspective, I do not see a significant speed increase. For the last 10 years, I have had an accuracy slider letting me choose between accuracy and speed. And the words got to the screen about as fast then as now. I have no doubt that there are orders of magnitude more processing and comparisons against ever larger dictionaries taking place behind the scenes nowadays. However, that is just the point: despite the the activity inside the black box, I do not see that much change with what happens on the screen.

I suspect most users who have been using voice recognition for more than a few years realize that it is as fully baked as it is gonna get. As much as the advertising focuses on accuracy and speed, I think it is other features that win over users. I think Dragon (Nuance) has been smart in adding functionality to Naturally Speaking (and now Dictate 2.0 for Mac) [Read more...]