VoxCommando
Microphones and Speech Recognition in General => Tips for better speech accuracy => Topic started by: vulcanjedi on March 15, 2014, 11:41:34 PM
-
Just saw a deal at Newegg for Dragon. Free.
Would this improve speech recognition if we can use their engine instead?
http://www.dealcatcher.com/redir/996757977
Nuance Dragon Naturally Speaking 12 Home
$0 $99.99 After Coupon & Rebate Newegg
Use Newegg coupon code EMCYTZT58823 during checkout. Newegg.com has Nuance Dragon Naturally Speaking 12 Home on sale for $64.99 after instant savings. Add the coupon code, and the price drops to free. This version of the popular text-to-speech software features improved accuracy, faster performance and an interactive tutorial. Shipping is $1.99.
-
I would have to agree that using D.N.S. engine instead of the Microsoft Speech Recognition engine would improve dictation and recognition drastically.
-
The DNS SDK licence runs a minimum of $5000.00 (five thousand) -- more, depending on the size of a developer's customer base, support needs, etc.
That, plus complete redevelopment of VoxCommando from the ground up is what would be required in order to offer a DNS-based version of VoxCommando.
Such a product would need to be priced much higher than the current, extremely affordable $40 licence cost for VC2. It would also require a fair amount of start-up capital to help support the hundreds of hours of development.
---
For people who need/wish to dictate essays to their computers, Dragon is the better engine, and anyone who wants to focus on dictation/free text should certainly look into getting themselves a home edition of Dragon dictation.
-
I would have to agree that using D.N.S. engine instead of the Microsoft Speech Recognition engine would improve dictation and recognition drastically.
I completely disagree. Dragon offers a better language model so it is certainly better for dictation, but I have done side by side tests and the raw accuracy of the engine was not noticeably different from what we are using now. For all our commands other than dictation there are many other factors which are much more significant. With a decent microphone and a correctly designed command structure I get very close to 100% accuracy. How can we possibly improve that drastically? If we improve from 99.5% to 99.7% is that drastically better?
Anyway, as Nime5ter points out, it isn't going to happen, but my point is that even if it were a realistic option, I don't believe it would be noticeably better, so people should not feel that they are somehow missing out. If dictation is what you are after then go buy Dragon, but that is not what VoxCommando is all about.
-
I completely disagree. Dragon offers a better language model so it is certainly better for dictation, but I have done side by side tests and the raw accuracy of the engine was not noticeably different from what we are using now. For all our commands other than dictation there are many other factors which are much more significant. With a decent microphone and a correctly designed command structure I get very close to 100% accuracy. How can we possibly improve that drastically? If we improve from 99.5% to 99.7% is that drastically better?
Anyway, as Nime5ter points out, it isn't going to happen, but my point is that even if it were a realistic option, I don't believe it would be noticeably better, so people should not feel that they are somehow missing out. If dictation is what you are after then go buy Dragon, but that is not what VoxCommando is all about.
I guess I will have to let go of my biased against MS speech rec and give it a legitimate chance. I have always used studio quality microphones so I dont usually experience bad dictation with DNS. I had no idea the sdk cost so much, wow. However at the end of the day it just boils down to the program working like it should, not who made the speech rec engine.