Off-Topic: Computing for the Blind

[Da Rock]

On 03/27/12 20:41, Jerry wrote:
> On Mon, 26 Mar 2012 18:50:06 -0400
> Robert Huff articulated:
>
>> Polytropon writes:
>>
>>>   Speech recognition requires training. Both the user and the
>>>   system have to learn from each other. But you have a learning
>>>   curve everywhere, be it typing, talking, or reading from a
>>>   Braille output.
>> 	In the case of speech recognition, that's a curve many might
>> be willing to travel if they had reason to believe it was effort
>> wisely invested.
>> 	There are a couple of ports that cleim to do speech
>> recognition.  Does anyone have experience with them?
> When it comes to "speech recognition", the only two applications that
> seem to work reliably at all levels are "Siri on iPhone 4S" and "Dragon
> NaturallySpeaking", neither of which are obviously available on
> FreeBSD. I don't believe that there is even a *nix/BSD version of
> "Dragon NaturallySpeaking" in production. In any case, I do have a
> friend who is severely vision impaired that uses that software with
> amazing results. She can definitely dictate a letter faster than I can
> manually create one.
The biggest contender in ports is sphinx- libraries are used as a basis 
for siri and the google offering. This is apparently used by phone 
companies, etc. Each of which use teams of developers to get it working 
the way they want. Getting it to work on an individual basis...

Apparently the results will primarily vary based on the dictionaries 
that are supplied, so it does mean one may work better than the other.