Secrets of VDct:,Replacing dictation components in Dragon NaturallySpeaking from Eric S. Johansson on 2010-09-10 (public-xg-htmlspeech@w3.org from September 2010)

From: Eric S. Johansson <esj@harvee.org>
Date: Thu, 09 Sep 2010 22:09:50 -0400
To: public-xg-htmlspeech@w3.org
Message-ID: <4C89936E.3080204@harvee.org>

  http://www.synapseadaptive.com/joel/VoiceCoders.ppt

Here's an example of how NaturallySpeaking does things under the covers. It's 
what I'd consider a fairly old model and it's been around for 10 years give or 
take. I'm not advocating it as a path to follow but an example of a working 
system with a significant history.

Take a look at slide 10. vdct overview. The concept of "hidden edit control" is 
quite an interesting one.  we have two mirror images of visible text region. The 
keyboard changes one, speech recognition changes the other. If one changes, it 
pushes its changes to the other.  This split was useful in the VR-mode because 
of the different representations of text in Emacs versus "normal" Windows.

Received on Friday, 10 September 2010 02:11:21 UTC