Newsgroups: comp.speech
Path: cantaloupe.srv.cs.cmu.edu!rochester!cornell!travelers.mail.cornell.edu!news.kei.com!news.mathworks.com!usenet.eel.ufl.edu!news-feed-1.peachnet.edu!gatech!howland.reston.ans.net!ix.netcom.com!netcom.com!ebohlman
From: ebohlman@netcom.com (Eric Bohlman)
Subject: Re: Text-to-speech as basic OS Service
Message-ID: <ebohlmanD4o4J4.B6F@netcom.com>
Organization: OMS Development
X-Newsreader: TIN [version 1.2 PL1]
References: <3inmk5$kc@hahn.informatik.hu-berlin.de> <D4nz3C.1DI@watserv3.uwaterloo.ca>
Date: Mon, 27 Feb 1995 16:59:27 GMT
Lines: 36
Sender: ebohlman@netcom13.netcom.com

Mark Stephen Pundsack (mspundsa@coulomb.uwaterloo.ca) wrote:

: Wouldn't many programs simultaniously trying to talk be distracting?  They'd 
: each have to have a different voice and make sure they don't speak out of
: turn.  Also, if the speech output just says something and then goes away,
: wouldn't if have a high chance of being missed?  A dialog box stays there
: until you acknowledge it was read.  

: Personally, I'd like to see an OS that supports speech output and input 
: in a seemless way.  If the speech output could be acknowledged by speech
: input, then it makes sense.  You have to set some thresholds for how much
: speech you want disrupting you.  I would find spontaneous speech to be 
: disruptive in my work.  I think you might have to go all the way to a
: co-worker metaphor.  Have a small icon of a person appear as if they are
: waiting to speak with you.  Or they can interrupt by knocking or saying,
: "Excuse me...".  When you acknowledge that you are ready to listen, they
: will continue.  Then you must acknowledge that you understand the message.
: This gets much more involved then simply replacing dialogs with speech.

Most definitely it's more involved than just "speak messages as they 
appear."  In fact, when blind users are evaluating screen reader (the 
software)/voice synthesizer (the hardware) combinations for accessing a 
computer, they usually give more weight to the ability to quickly shut 
the speech up when they've heard what they want and move to something 
else than they do to any other characteristic (if you're listening to a 
menu and hear the option you want, you don't want to have to wait for the 
rest of the menu being read before you can proceed).  Also, one of the 
major functions of a screen reader is to let the user control what he/she 
wants to hear, and when.

Speech output from a computer has to be conversational (regardless of 
whether the user's part of the conversation is accomplished by speech 
input or other means); simply having the computer talk at the user (and 
program-directed speech output without user control is talking at the 
user rather than to the user) doesn't accomplish much.

