- From: Eric S. Johansson <esj@harvee.org>
- Date: Wed, 08 Sep 2010 10:55:45 -0400
- To: Satish Sampath <satish@google.com>
- CC: Olli@pettay.fi, Dan Burnett <dburnett@voxeo.com>, public-xg-htmlspeech@w3.org
On 9/8/2010 10:26 AM, Satish Sampath wrote: > Hi Olli, I've been reading a bit of the background material and I admit I still to read a bit more. My first question is How many of you live with speech recognition like I do? I should be living without a keyboard but, way too much of the world is hostile to those of us with a upper extremity disorders. I'm also wondering who is the target audience for making these changes? Is it tabs or, can the disabled take care of themselves? XML is one of the worst formats for data if you are disabled. It is something you would never try to speak using speech recognition simply because of the hard to pronounce character sequences, the requirement that you speak of one character at a time usually discreet utterance form to make sure you get the right recognition which is tantamount to speaking the keyboard. Speaking the keyboard is a well known user mode imposed upon the disabled with the end result that the disabled lose the ability to speak because it is so wearing on the throat. This may be because I haven't read the background yet but how do you accommodate for editing? I'm sure I'll leave several speakos behind in this message for no other reason than it's impossible to crack by speech and my hands are burned out enough I don't want to bother navigating. I should probably explain my background. I've been disabled for over 15 years, speech recognition user the entire time. Watched the industry progress from discrete utterance to large vocabulary continuous recognition systems. I was part of the crew hoping to put together and organize the first generation of programming by voice systems. I organized the conference is and lined up talent to talk to us. I was part of the now defunct open source speech recognition initiative program and I'm now trying desperately to figure out enough about natpython so that I can apply Select-and-Say features to pyscripter. Quite a challenge when your hands aren't functional. Obviously I've learned quite a lot about what it's like to live with speech. I've learned that simple macro models are a dead-end. They're only good for trivial stuff. most "command-and-control interfaces are also dead ends. For the most part, he was my hands as broken as they are, I don't speak anything for command-and-control. It's too stressful to the throat. similarly, "natural language" commands are pretty much useless. Too hard to discover, dangerous with misrecognition's, set wrong user expectations What you need to look at is creating "written speech" interfaces for more complicated operations. I forget the actual terms but a written speech interface is one where you can dictate into some holding editor and then translate the dictated information to what you need. It gives you opportunities for correction you wouldn't otherwise have. It also lets you pause and think in the middle of an utterance. For example, you could be spelling out a change directory command. You forget where you are in the hierarchy so when you pause, the system comes up with a prompt tell you what can be said and then you say it through to the end. While the command is in the miniature editor, you can change individual pathnames (because they are all spoken words), and even use additional commands like "remember this as <register name>" to store or retrieve the path. Anyway, that's the direction I'm heading down with my speech user interface ideas. I'm hoping that I can nudge you down the path of making something usable by crips like me. I highly recommend you throw away your keyboards and use NaturallySpeaking for everything. It's the current gold standard. It's quite enlightening.
Received on Wednesday, 8 September 2010 14:57:53 UTC