See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/2334517
Matching a tone-based and tune-based approach to English intonation for
concept-to-speech generation
Article · July 2000
DOI: 10.3115/992730.992766 · Source: CiteSeer
CITATIONS READS
0 17
2 authors:
Elke Teich Catherine I Watson
Universität des Saarlandes University of Auckland
83 PUBLICATIONS 443 CITATIONS 71 PUBLICATIONS 791 CITATIONS
SEE PROFILE SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Speech Forensics View project
Sound Change in New Zealand English View project
All content following this page was uploaded by Catherine I Watson on 09 September 2014.
The user has requested enhancement of the downloaded file.
, Matching a tone-based and tune-based approach to English
intonation for concept-to-speech generation
Elke Teich
Universitgt des Saarlandes, Saarbr{icken & University of Sydney
C a t h e r i n e I. W a t s o n and C d c i l e P e r e i r a
Macquarie University, Sydney
Abstract tion is one of the major factors (ct'. Cole et
Tlle paper describes the results of a compari- al. (1995)). The concrete goal we are pursu-
son of two annotation systems for isstoslal;ion, ing is to connect an oil-the-shelf speech syn-
the tone-based ToBI al)proach and the 1;une- thesizer for English (FESTIVAL; (Black et al.,
based api)roach proposed by Systemic Func- 1998)) with an automatic text generation sys-
ti(mal Grammar (SFO). The goal of this compar- tem tbr English based on SFO (Matthiessen &
ison is to detine a mapping between the two sys- Bateman, 19911. Since in the SFO approach, in-
tems tbr the purpose of concept-to-speech gen- tonation is accounted for as part of grammar
eration of English. Since ToB: is widely used rather than as an independent component, it is
in Sl)eech synthesis and SFG is widely used in straightforward to extend the grammatical re-
nal;ural language generation and oft~rs a lin- sources of a systemically based text generation
guistically motivated aecollnt of intonation, it; system with an account of intonation (cf Teich
appears a promising step to comt)ine the two et al. (1.997) iml)lenmnting such all approach for
approaches for concept-to-speech. A corpus of German concet/t-to-speech generation). Con-
English utterances has been analysed with both necting such a system to a speech synthesizer
~].~()13I and SFG categories; eomparison of the requires mapping the OUtl)ut of the generator
analysis results has lead to the identification of to the input requirements of the st)eech synth(>
some basic equivalents between the two systems sizer. In the FESTIVALsystei11, the intonation of
on which a mapping can be based. the text to be synthesized can be manipulated
1)y ~mnotation with TOBI labels. Therefore, a
1 Introduction mapl)ing betweeIl the SFC and the ToBI anno-
The pallet describes the main results of a con> tation systems is required.
parison of /;he ToB: (Tone-and-Break-Indices) Second, there is a theoretical lnotivation.
ai)proach (Pierrehumbert, 1.9801 Silverman el; With a mapping between tile ToBI and the slpo
al.., 19961 to annotating English speech data systems for intonation almotation, it will be
with information about intonation and one of possible to link the 1)honetic analysis of speech
the British School approaches (e.g., Brazil et al. data to an interpretation of intonational mean-
(1980)), Systenfie Fmmtional Grammar (SFO; ing as it is proposed by SFO. Existing speech
(Halliday, 19671 Halliday, 1970)). The goal of corpora that are acoustically analysed and an-
this comparison is the definition of a mapping notated with ToBI tail then be used to test
between the two systems. some of the assumptions brought forward by
This attempt has a two-fbld motiw~tion. SFO about the natm:e of intonation. Also, with
First, it is motivated by computational appli- a mapping between ~oBI and SFG annotations,
cation in concept-to-si)eech systems, in which an exchange of annotated corpora between ToBI
text in spoken mode is automatically generated and SFO users would be possible.
from an underlying abstract lneaning represen- We report on the analysis of a sl)eech cor-
tation, it is widely acknowledged that in order pus compiled fl'om Halliday (1970) with ToBI
for spoken language technology to gain wider and SFO labels (See. 3). The intonation analy-
acceptance, it has to improve on the quality of sis is based on an acoustic analysis of the speech
output considerably. Itere, appropriate intona- data in terms of fundamental frequency (F0).
829
Matching a tone-based and tune-based approach to English intonation for
concept-to-speech generation
Article · July 2000
DOI: 10.3115/992730.992766 · Source: CiteSeer
CITATIONS READS
0 17
2 authors:
Elke Teich Catherine I Watson
Universität des Saarlandes University of Auckland
83 PUBLICATIONS 443 CITATIONS 71 PUBLICATIONS 791 CITATIONS
SEE PROFILE SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Speech Forensics View project
Sound Change in New Zealand English View project
All content following this page was uploaded by Catherine I Watson on 09 September 2014.
The user has requested enhancement of the downloaded file.
, Matching a tone-based and tune-based approach to English
intonation for concept-to-speech generation
Elke Teich
Universitgt des Saarlandes, Saarbr{icken & University of Sydney
C a t h e r i n e I. W a t s o n and C d c i l e P e r e i r a
Macquarie University, Sydney
Abstract tion is one of the major factors (ct'. Cole et
Tlle paper describes the results of a compari- al. (1995)). The concrete goal we are pursu-
son of two annotation systems for isstoslal;ion, ing is to connect an oil-the-shelf speech syn-
the tone-based ToBI al)proach and the 1;une- thesizer for English (FESTIVAL; (Black et al.,
based api)roach proposed by Systemic Func- 1998)) with an automatic text generation sys-
ti(mal Grammar (SFO). The goal of this compar- tem tbr English based on SFO (Matthiessen &
ison is to detine a mapping between the two sys- Bateman, 19911. Since in the SFO approach, in-
tems tbr the purpose of concept-to-speech gen- tonation is accounted for as part of grammar
eration of English. Since ToB: is widely used rather than as an independent component, it is
in Sl)eech synthesis and SFG is widely used in straightforward to extend the grammatical re-
nal;ural language generation and oft~rs a lin- sources of a systemically based text generation
guistically motivated aecollnt of intonation, it; system with an account of intonation (cf Teich
appears a promising step to comt)ine the two et al. (1.997) iml)lenmnting such all approach for
approaches for concept-to-speech. A corpus of German concet/t-to-speech generation). Con-
English utterances has been analysed with both necting such a system to a speech synthesizer
~].~()13I and SFG categories; eomparison of the requires mapping the OUtl)ut of the generator
analysis results has lead to the identification of to the input requirements of the st)eech synth(>
some basic equivalents between the two systems sizer. In the FESTIVALsystei11, the intonation of
on which a mapping can be based. the text to be synthesized can be manipulated
1)y ~mnotation with TOBI labels. Therefore, a
1 Introduction mapl)ing betweeIl the SFC and the ToBI anno-
The pallet describes the main results of a con> tation systems is required.
parison of /;he ToB: (Tone-and-Break-Indices) Second, there is a theoretical lnotivation.
ai)proach (Pierrehumbert, 1.9801 Silverman el; With a mapping between tile ToBI and the slpo
al.., 19961 to annotating English speech data systems for intonation almotation, it will be
with information about intonation and one of possible to link the 1)honetic analysis of speech
the British School approaches (e.g., Brazil et al. data to an interpretation of intonational mean-
(1980)), Systenfie Fmmtional Grammar (SFO; ing as it is proposed by SFO. Existing speech
(Halliday, 19671 Halliday, 1970)). The goal of corpora that are acoustically analysed and an-
this comparison is the definition of a mapping notated with ToBI tail then be used to test
between the two systems. some of the assumptions brought forward by
This attempt has a two-fbld motiw~tion. SFO about the natm:e of intonation. Also, with
First, it is motivated by computational appli- a mapping between ~oBI and SFG annotations,
cation in concept-to-si)eech systems, in which an exchange of annotated corpora between ToBI
text in spoken mode is automatically generated and SFO users would be possible.
from an underlying abstract lneaning represen- We report on the analysis of a sl)eech cor-
tation, it is widely acknowledged that in order pus compiled fl'om Halliday (1970) with ToBI
for spoken language technology to gain wider and SFO labels (See. 3). The intonation analy-
acceptance, it has to improve on the quality of sis is based on an acoustic analysis of the speech
output considerably. Itere, appropriate intona- data in terms of fundamental frequency (F0).
829
