[Kde-accessibility] Simon

[Peter Grasch]

Hi Steve,

On Saturday, May 17, 2014 06:03:23 PM you wrote:
> Well I'd love to see dictation in Simon.  Is it really only the lack of
> a speech corpus that makes it not so?
No, of course not. To build a dictation capable system, you need a good 
acoustic model, a good language model and a dictating application. We have 
starting points for all of them but for a serious system, they all need work.

> There is a very active Brazilian group at Coruja which I believe is
> linked to the FalaBrasil group at
> http://www.laps.ufpa.br/falabrasil/simonbr.php.  As far as European
> Portuguese is concerned, I'm not aware of similar groups.
> 
> Do these groups not have the required vocab?
No, that's by far not the biggest problem, it's relatively easy to get 
relatively good vocabulary lists (crawling online resources to get a text 
corpus and using simple counting to select the N most frequent words is a 
common and good strategy).

> I guess from your reply, that it's not available in English or German
> either.
Correct.

> For our command and control we only use about 5 commands and we have
> sampled these independently.  Amazingly they even work when I use them
> in my strong British accent.
Sure. But the more choice, the more difficult.

> In the future we would like to add a list of medical words, but in the
> meantime we'll use some form of pointers, eg popups controlled
> alphabetically.
Good idea. However, medical terminology is actually very easy to recognize 
(distinguish) as the words are so long. Coming back to dictation, "pulmonary 
edema" is actually arguably easier to write than "hey" - and definitely easier 
than "hay" :)

Best regards,
Peter