[MV] IBM ViaVoice 3 and 10.4.7

[Chuck Rogers]

Henry (and everyone else):

It has to do with how development resources have been allocated in  
the respective companies (Dragon and MacSpeech).

Dragon has had the benefit of having a codebase virtually unchanged  
since Windows 95. This has given them the last 11 years plus to  
improve their algorithms, add features, and just in general, tweak  
their software.

MacSpeech, on the other hand, has had to rewrite our software from  
the ground up three times just to keep it running on the Mac: once to  
go from 68K to PPC, again to go from Mac OS 9 to Mac OS X, and just  
recently, yet again to go from PPC to Intel.

We have our fingers crossed that Apple's "top secret" features for  
10.5 will not require us to further retool our software in any  
substantial way. But the bottom line is that while Dragon has been  
able to significantly improve their training process, we have been  
mired in just keeping our software running on the platform.

The good news is that all speech recognition software out there uses  
pretty much the same methodology: Hidden Markov Modeling. That's why  
you can eventually get accuracy from iListen on par with Dragon - it  
just takes longer to get there with iListen.



Best Regards,

Chuck Rogers, Chief Evangelist
MacSpeech, Inc.




On Sep 5, 2006, at 5:14 PM, Henry Kalir wrote:

> Hi Chuck,
>
> I'm considering a move to the most recent iListen. Just curious -  
> why would iListen take longer than Dragon to achieve the same  
> accuracy?
>
> Best,
>
> Henry
>
> Chuck Rogers wrote:
>