Re: text-to-speech and standards

     
     >Jim, is there anything up on the web that would pass for a situation 
     >survey?  Something about what is in practice in the field, and 
     >possible opportunities to improve things by defining standards?
     
     I'm not sure, because there are several directions to explore.  
     
     1. What is the current state of the art in text-to-speech technology?
     
        This is not as nice as people would like, because the humanlike 
     quality falls short of what is desired.  Our speech synthesizer 
     interface library ("SSIL"), the leading standard used in applications 
     developed for the blind, is a good example of the state of the art 
     five years ago.  Unfortunately, the state of the art in TTS hasn't 
     changed much except for the need to live in 32 bit environments!
     
     2. What should we plan for in the future?     
     
        My impression from reading the briefing package is that we're 
     discussing aiming quite high, trying to mark up material with very 
     rich information.  The question also exists of where the primary 
     responsibility for natural speech output lies: in the document creator 
     or the output device.
        Are we aiming for the least common denominator browser client in 
     content or a richer one?  The answer is usually yes, both.
        As someone who is mainly interested in developing talking client 
     applications, I'd like to be able to deliver all of what is possible.  
     However, that's a high bar!
     
     3.  What should a talking browser do in handling different kinds of 
     objects?
     
        Some applications call out different objects with a "LINK" 
     statement.  I agree with you that a slight change in the TTS would be 
     desirable for some users.  I've done an application for reading 
     scanned material aloud that has five different voices (reading, menu, 
     bold, italic and underline), and think that this would be a valuable 
     way to convey information.  I think the underlying information should 
     be there and each user make the choice of how it is conveyed.
     
     I'm just feeling my way in this group, trying to figure out where 
     we're going technically.  
     
  Jim Fruchterman                    jim@arkenstone.org
  President                          Arkenstone, Inc.
  555 Oakmead Parkway                1-800-444-4443
  Sunnyvale, CA 94086 USA            1-408-245-5900
  "Information Access for Everyone!" Fax: 1-408-328-8484
  http://www.arkenstone.org

Received on Tuesday, 13 May 1997 09:02:35 UTC