Wokshop on developing an International Prosodic Alphabet (IPrA) within the AM framework Sun Sun Sun Sun- - -Ah Jun Ah Jun Ah Jun Ah Jun, UCLA Jos Jos Jos José é é Ignacio Hualde Ignacio Hualde Ignacio Hualde Ignacio Hualde, University of Illinois at Urbana-Champaign Pilar Prieto Pilar Prieto Pilar Prieto Pilar Prieto, ICREA – Universitat Pompeu Fabra
73
Embed
Wokshopon developing an International Prosodic Alphabet ...linguistics.ucla.edu/ipra_workshop/IPRA_PRESENTATION2015.pdf · Wokshopon developing an International Prosodic Alphabet
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Wokshop on developing an International
Prosodic Alphabet (IPrA) within the AM
framework
SunSunSunSun----Ah JunAh JunAh JunAh Jun, UCLA
JosJosJosJoséééé Ignacio HualdeIgnacio HualdeIgnacio HualdeIgnacio Hualde, University of Illinois at
1. Introduction & the motivations for 1. Introduction & the motivations for 1. Introduction & the motivations for 1. Introduction & the motivations for
developing the IPrAdeveloping the IPrAdeveloping the IPrAdeveloping the IPrA (Jun, Hualde, Prieto)
2. Proposals on labels of2. Proposals on labels of2. Proposals on labels of2. Proposals on labels of---- Pitch accents (Prieto)Pitch accents (Prieto)Pitch accents (Prieto)Pitch accents (Prieto)---- Phrasal/Boundary tones (Jun)Phrasal/Boundary tones (Jun)Phrasal/Boundary tones (Jun)Phrasal/Boundary tones (Jun)---- NonNonNonNon----f0 features (Hualde)f0 features (Hualde)f0 features (Hualde)f0 features (Hualde)
Part 1Part 1Part 1Part 1
Introduction & the motivations Introduction & the motivations Introduction & the motivations Introduction & the motivations for developing the IPrAfor developing the IPrAfor developing the IPrAfor developing the IPrA
(by Sun-Ah Jun)
Transcription system of intonation and prosodic
structure in AM framework
• ToBI (Tones and Break Indices) is a consensus system for labelling spoken
utterances to mark phonologically contrastive intonational events and
prosodic structure based on the Autosegmental-Metrical model of intonational
Proposal: Develop an International Prosodic Alphabet (IPrA)
NOTIFICA
Develop a set of discrete tonal labels and diacritics that are
transparent and consistent at the categorical phonetic level.
This will be used:
1. as a temporary label before establishing a phonological
analysis of tones
2. as a way to represent allophonic realizations of an underlying
tonal category
3. as a way to represent hybrid or exceptional tonal categories
that are not part of the intonational model of any specific
language.
DO WE NEED A UNIVERSAL SET OF
PROSODIC LABELS?
(by Jose Hualde)
Comparative work
Difficult or impossible if labels have different
interpretations in different languages or
analyses
Current situation
The same contour may be given different labels
The same label is used for different contours
Examples: final sustained pitch and
final rise
Advantages of a common understanding
of symbols
Answering typological questions such as:
How many stress languages have a contour with a fall from
the pretonic to the stressed syllables?
H+L* (or HL*?)
What is the distribution/pragmatics of this contour in
different languages?
L*+HENGLISH
e.g. Veilleux et al. (2008)
L*+HSPANISH
e.g. Beckman et al (2002)
L+H*SPANISH
e.g., Prieto and Roseano (2010)
L+H*GREEK
e.g., Arvaniti and Baltazani (2005)
Different interpretations of same label:
L*+H and L+H* in different ToBI systems
Another question:
Why not use IPA tone diacritics?
Syllable-by-syllable tonal transcription systems
make generalizations across utterances
difficult
Autosegmental labels capture the relation
between underlying/phonological and
broad/categorical phonetic levels of
description in a more conspicuous way.
Autosegmental notation was introduced for the analysis
of lexical tone in order to better account for the
mapping between the broad phonetic level and the
postulated phonological level, including phenomena
such a contour formation from underlying sequences
of tone, tone spreading, surfacing of tone on different
syllables from their lexical sponsor, floating tones,
etc.
Bruce (1977) demonstrated the usefulness of the
autosegmental approach in our understanding of the
intonational contours of Swedish, by providing a uniform
underlying representation for the two contrastive lexical
pitch-accent, in spite of surface variation as lexical and
postlexical tones interact.
A clear advantage of the AM symbols is in
indicating differences of alignment between
segments and tonal events.
Also:
IPA symbols for “global rise” and “global fall” (in
addition to lexical tone and word accent ) do not
appear to be enough to capture all relevant
facts in intonation.
WHY DO WE NEED TWO LEVELS OF PROSODIC TRANSCRIPTION?
(by Pilar Prieto)
Main argument: Because it is very useful to represent the correspondence between
underlying prosodic categories and surface
patterns
In this section, we provide a set of examples (and more arguments) to motivate the need for a two-level approach to prosodic annotation
Broad phonetic transcriptions,
segmental level
• Broad phonetic transcriptions Broad phonetic transcriptions Broad phonetic transcriptions Broad phonetic transcriptions are very commonly used at the segmental level.
• They include easily heard characteristics and ignore important phonetic detail.*
Broad phoneticBroad phoneticBroad phoneticBroad phonetic meter (EnglishEnglishEnglishEnglish) mira ‘s/he looks’ (CatalanCatalanCatalanCatalan)
[ˈmiɾə] [ˈmiɾə]
• Similar broad phonetic transcriptions, yet different phonological analyses phonological analyses phonological analyses phonological analyses :
PhonologicalPhonologicalPhonologicalPhonological meter (EnglishEnglishEnglishEnglish) mira ‘s/he looks’ (CatalanCatalanCatalanCatalan)
/ˈmittttər/ /ˈmiɾ+aaaa/
* A narrow phonetic transcription narrow phonetic transcription narrow phonetic transcription narrow phonetic transcription would encode finer differences of phonetic detail.
• Broad phonetic transcriptions are useful for the systematic description of
the phonetic realization of underlying segments.
• For example, in Spanish, it is important to know that [ˈehta] and [ˈesta] are
two ways of pronouncing the same word esta /ˈesta/ ‘this, fem.’ in many
dialects.
Conversely, having access ONLY to a broad phonetic level
would not inform us about the fact that they are phonetic
realizations of the SAME category.
Having access ONLY to a phonological level of transcription
would not inform us about the surface patterns of syll-final /-s/.
These two levels of representation have been the
standard way of analyzing sound alternations across
languages.
The same arguments can be applied at the suprasegmentalsuprasegmentalsuprasegmentalsuprasegmental
levellevellevellevel.
In here, we will consider two types of examples which
• Truncated tunes are quite frequent crosslinguistically
• It is quite common crosslinguistically for a rising-falling underlying intonation
sequence such as L+H* L% to be truncated if lexical stress falls on the word-final
syllable (see Grice et al. 2005 for southern Italian varieties, Ortega-Llebaria and Prieto 2009 for
Catalan and Peninsular Spanish, Armstrong 2010 for Puerto Rican Spanish, Gabriel et al. 2010 for
Argentinean Spanish, and Cabrera-Abreu and Vizcaino-Ortega 2010 for Canarian, among many others)
• Yet, current ToBI practices make it hard for researchers to systematically refer to
truncated patterns.
Truncated interrogative tunes, Puerto
Rican Spanish
• “The phonetic realization of a ¡H* L% interrogative tune can initially cause
confusion for transcribers” (Armstrong 2015).
• In stress-final words it can be truncated, while in paroxytonic words it is not.
¿Que vieron a Mariariariariana?‘Did you see Marianne?’
¿Se quieren callarllarllarllar?‘Do you want to stop talking?’
Transcribing truncated contours
• Having access ONLY to a phonological level of transcription does not allow
researchers to systematically refer to a representation of truncated patterns.
• Conversely, having access ONLY to a broad phonetic level of transcription
does not inform us about the fact that we are dealing with surface realizations
of the SAME category.
• Contextual neutralization Contextual neutralization Contextual neutralization Contextual neutralization is a pervasive phenomenon in segmental phonology
and, arguably, its incidence is very strong in the intonational component.
• Not allowing for a level of broad phonetic representation means that we will
not have access to the level of representation level of representation level of representation level of representation that represents allophonic allophonic allophonic allophonic
differencesdifferencesdifferencesdifferences (e.g., distinguishes between truncated vs. non-truncated contours,
etc.).
Neutralization of contrasts: L+H* and H*
• The potential contrast between H* H* H* H* and L+H* L+H* L+H* L+H* has been the source of a good
amount of interinterinterinter----transcriber disagreement transcriber disagreement transcriber disagreement transcriber disagreement across several ToBI systems (e.g.
Pitrelli et al 1994 and Syrdal et al 2001 for Mainstream American English ToBI, and
Escudero et al. 2012 for Catalan ToBI).
• For American English ToBI the issue has not been settled yet (e.g. work by Ladd
2008a, Ladd and Morton 1997 arguing for a gradient difference, Steedman 2013 for a
categorical difference, and also depends on dialect).
• However, phrase-initial positions are a clear neutralizing context in English
(MAE_ToBI) and in other languages with the L+H*/H* alternation.
Catalan neutralization of L+H* vs. H*
In Catalan, we have a clear contrast between a H* and a L+H* pitch accent.
And, like in English, phrase-initial L+H* can surface as H*.
• Again, if we are interested in analyzing the patterns of correspondence
between the underlying L+H* pitch accent and its surface realizations, one
level of analysis is not sufficient.
• Some work has adopted an intermediate level of analysis, like (L)+H* (e.g.,
Grice 2005). Yet, in a system with phonological H* (Catalan), a H*
transcription can ambiguously correspond to two distinct phonological
categories.
ADVANTAGES OF ADOPTING TWO LEVELS OF PROSODIC TRANSCRIPTION
1.It allows for a clearclearclearclear mapping between underlying categories of prosody mapping between underlying categories of prosody mapping between underlying categories of prosody mapping between underlying categories of prosody and their surface patterns.their surface patterns.their surface patterns.their surface patterns.
And…
2.It can allow for more abstract phonological analyses more abstract phonological analyses more abstract phonological analyses more abstract phonological analyses of intonation across languages.
3.It can contextualize results of detailed phonetic analyses contextualize results of detailed phonetic analyses contextualize results of detailed phonetic analyses contextualize results of detailed phonetic analyses of tunes.
Abstract phonological analyses
• Another advantage of the two-level analysis would be the clarification of the
status of each level of transcription.
• Some ToBI systems represent a compromise between broad phonetic and
phonological levels of transcription. E.g., Korean ToBI and French ToBI
examples.
• In Spanish ToBI, the label L+<H* is very useful as a broad prosodic label to
analyze dialectal differences (e.g., retracted peaks reported in Andean Spanish by
O’Rourke 2005, Buenos Aires Spanish by Colantoni and Gurlekian 2004; Spanish in
contact with Basque by Elordieta and Calleja 2005), but no phonological differences
in alignment are reported in prenuclear positions.
• If transparent labellings are encoded at the broad phonetic level,
important generalizationsgeneralizationsgeneralizationsgeneralizations and nonnonnonnon----predictable information predictable information predictable information predictable information can be
encoded at the phonological level.
• Conversely, adopting a level of broad phonetic transcription will also help
clarify systematic analyses of fine phonetic features systematic analyses of fine phonetic features systematic analyses of fine phonetic features systematic analyses of fine phonetic features across intonation
contours (F0 Max, F0 Min, duration, etc.), which will be able to incorporate a
level of broad phonetic transcription.
• Oftentimes, this work does not incorporate a phonological analysis of the
pitch contours and uses general terms such as “global rise”, “global falling”,
etc.
• Having access to an IPrA alphabet to transcribe the curve will facilitate the
use of the tools for broad phonetic transcription.
• We regard this proposal as an opportunity to integrate
purely phonetic vs. phonological work on intonation
PARTIAL CONCLUSIONS
IPrA PROPOSAL IN A NUTSHELL
•Adoption of two levels of analysis, e.g. broad phonetic and phonological.
• Arguments for developing an IPrA alphabet which is based on units that are universally accepted.
Advantages of having broad phonetic transcription in
addition to the phonological tones in ToBI
• To facilitate the prosodic labelling of languages for which the phonological
repertoire is not yet known.
• To allow for more transparent comparisons across languages.
• To clarify in a systematic way the relationship between the phonetic forms and
phonological categories.
• To clarify the level of transcription different ToBI systems are using right now.
• To allow for more abstract phonological analyses of intonation across languages.
• To facilitate the use of tools for broad phonetic transcription to researchers
interested in fine phonetic detail.
• To increase the levels obtained by inter-transcriber agreement tests.
• To facilitate the study of languages in contact and L2 prosody.
• To facilitate the automatic and semi-automatic labelling of large corpora.
Part 2
Brief practical proposal and rationale behind the IPrA set
Rationale behind the construction
of an IPrA set
NOTIFICA
• Proposal based on prosodic descriptions of more than 30 languages (e.g., languages included in Jun 2005, Jun 2014, Frota & Prieto 2015, among others).
• New units are incorporated if they can be shown to be used with a distinctive value in a given language. That is, we are adopting a tone symbol for the broad phonetic representation if the symbol was used in some language to have a distinctive value.
• This is similar in nature in how the IPA incorporates a new symbol in the segmental inventory.
• Our goal is to establish a set of “broad phonetic tonal labels”, that is, we are trying to capture the categorical nature of f0 contour, thus not narrow phonetic representation of f0 (like INTSINT and IViE),
• In a new language, where there is no clear presence of lexical stress or pitch accent, one could use temporary labels without the ‘*’ or the ‘%’markers.
Pitch accents
Monotonal pitch accents
!H* at the broad phonetic level would be transcribing both phonological and non-phonological downstep. The phonological level of analysis would just encode the phonological level.
The H* and L* categories have been vastly used. Also, a handful of studies propose a contrastive mid tone, e.g. !H* (e.g., Beckman & Hirschberg 1994, Ipek & Jun 2013, Ipek 2015), and a super-high tone, i.e., ¡H* (Prieto 2014).
Bitonal accents: alignment contrasts
QUESTION: Do we have any evidence for a three-way phonological contrast in alignment for falling accents?
Bitonal accents: pitch height contrasts
QUESTION: Do we have evidence for a three-way scaling contrast in rising pitch accents (e.g., L+H* vs L+¡H* vs. L+!H*).
Tritonal pitch accents
QUESTION: Do we have evidence for contrastive tritonal pitch accents?
Phrasal/Boundary tones
Boundary tones
Monotonal:
IP-final boundary tones: low: L%high: H%super-high: ¡H% (vs. ^H in German ToBI)med high: !H% (vs. M% in Spanish, Jamaican Eng Creole)
high plateau: H-% (in German ToBI; H-L% in MAE_ToBI; H:% in
Cantonese; 0% in German by Grabe1998)mid plateau: !H-% (in German ToBI; % in Dutch for continuing
raised pitch range of a whole IP: %H-> (vs. %H which is local)
pitch range modification within an IP (all from Pan-Mandarin)beginning of a raised pitch range: %q-raisebeginning of local expansion of pitch range due to
emphatic prominence: %e-prom
beginning of pitch range reduction after focus: %compressed
Phrasal/Boundary tones
Intermediate Phrase (ip) boundary tones or phrase accent: