Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84; site utcsri.UUCP
Path: utzoo!utcsri!flaps
From: flaps@utcsri.UUCP (Alan J Rosenthal)
Newsgroups: comp.sys.amiga
Subject: Re: Questions
Message-ID: <4088@utcsri.UUCP>
Date: Sat, 7-Feb-87 18:08:58 EST
Article-I.D.: utcsri.4088
Posted: Sat Feb  7 18:08:58 1987
Date-Received: Sun, 8-Feb-87 06:44:41 EST
References: <775@vaxb.calgary.UUCP> <34@f.gp.cs.cmu.edu>
Reply-To: flaps@utcsri.UUCP (Alan J Rosenthal)
Organization: University of Toronto
Lines: 41
Summary: 


In article <34@f.gp.cs.cmu.edu> mjw@f.gp.cs.cmu.edu.UUCP writes:
>2)The narrator device is very nasty. I had this neat idea, "I wonder if I
>can get this thing to speak french" I asked myself. So, I looked in the
>manual, and it talked about ipa and nice things like that. "Neat" I thought, 
>"a piece of cake". Then I looked closely. What it provides is something
>called arpabet. Not only doesn't it provide the phones I need to generate
>French... [some text deleted by ajr]
>
>I would like to suggest that the next release of intuition support a
>narrator device with a full set of IPA phones.

This would not be nearly sufficient.  The problem is not just a lack of
phonemes; the narrator speaks English (of a very particular variety!).  It
knows about intonation and relative stress levels, for example.  Adding more
phonemes would not be sufficient.  This is the fundamental difference between
phonemes and phones.  The narrator device takes phonemes, which are a
description of the sounds in an utterance relative to a given language and
only specified as well as needed to distinguish them from other sounds in the
same language.

What would really be needed is for the narrator to output data which is not
language-specific which another device could speak, such as perhaps the audio
device.  Then there would be three stages for language creation:

	Translator - English text to phonemes (not used by most serious
					       programs)
	Narrator - phonemes to phones (used by nearly any <whatever dialect
				       that is> English program)
	Audio - actual speech production (used by absolutely all programs)

ajr (in Linguistics as well as cs!)
-- 

Alan J Rosenthal

UUCP: {backbone}!seismo!mnetor!utcs!flaps, ubc-vision!utai!utcs!flaps,
      or utzoo!utcs!flaps (among other possibilities)
ARPA: flaps@csri.toronto.edu
CSNET: flaps@toronto
BITNET: flaps at utorgpu