Newsgroups: comp.speech
Path: pavo.csi.cam.ac.uk!warwick!pipex!uunet!decwrl!decwrl!koriel!sh.wide!wnoc-kyo!atrwide!atr-la!paul
From: paul@itl.atr.co.jp (Paul Taylor)
Subject: Re: Intonation discrimination
In-Reply-To: andrewh@mimic.ee.su.oz.au's message of Sun, 11 Jul 1993 23:57:58 GMT
Message-ID: <PAUL.93Jul12154839@as53.itl.atr.co.jp>
Sender: news@itl.atr.co.jp (USENET News System)
Nntp-Posting-Host: as53
Organization: /NFS/as51/users/paul/.organization
References: <1993Jul9.070302.1142@trl.oz.au> <21jv45INN7eh@jhunix.hcf.jhu.edu>
	<1993Jul11.235758.13166@ucc.su.OZ.AU>
Date: Mon, 12 Jul 1993 06:48:39 GMT
Lines: 56

In article <1993Jul11.235758.13166@ucc.su.OZ.AU> andrewh@mimic.ee.su.oz.au (Andrew Hunt) writes:

   In article <21jv45INN7eh@jhunix.hcf.jhu.edu>, wliu@hns.com (Weimin Liu) writes:
>> In article <1993Jul9.070302.1142@trl.oz.au> wwen@tardis.trl.OZ.AU
> > (Wilson X. Wen) writes:
> > >Could anyone on the net please tell me how to discriminate different
> > >tones (in a 5-tone system of intonation, for example) in speech wave
> > >(LPC/Cepstrum/Raw wave)?
> > 
> > Tones and intonation are different concepts.  Tones are about small
> > phonetic elements such as syllables, whereas intonation is about
> > phrases or sentences.  In tonal languages, such as Chinese, tones are
> > segmental (they make a difference in the meaning).  Intonation, on the
> > other hand, usually only indicates whether a sentence is a question or
> > a command, etc.
This is a bit of a simplification really. Tonal languages and Intonation
languages differ in that in tone languages pitch serves to distinguish
lexical items and therefore tonal patterns are marked in the lexicon. This
is somewhat similar to stress in English, although there arent
many cases of  lexical items differing in stress alone. (But there
are common examples such as "record" the verb and "record" the noun).

In Intonation languages, pitch is not defined in the lexicon, and the
speaker is more or less free to put any intonation pattern
on the utterance. In a language like English, intonation has many functions,
but typically you can say it is made up of "pitch accents" which vary
in type and are associated with single syllables; "phrasing" which 
indicates the prosodic structure of the utterance; and "pitch range"
which indicates the overall F0 range of a phrase etc. All these
effects contain "meaning", and intonation does much more than
signal question/statement; it indicates a variety of speech acts and
can be used to indicate constrastive stress and emotion.

>
> There is also a theoretical framework of intonational analysis developed
> by M. Halliday which uses the term "tone" to describe phrase level intonation
> patterns.  For example, tone 1 is a falling tone over a word or series of
> words.  The system consists of 5 primary tones, secondary tones and more, 
> so it seems that Wilson is using the correct terminology within that theory.

Halliday's terminology is very non-standard and his
work is much better understood by replacing "tone" with
"pitch accent" or "tune". His tones are really F0 patterns (configurations).
Pierrehumbert's theory uses the term "tone" in a manner much more
consistent with the use of the word in tone languages.

I can give more references on intonation labelling/recognition/
discrimination etc if you want.

Paul 

ATR, Japan
paul@itl.atr.co.jp



