barcelona listeners auditory impressions from a perceptive ...prosodia.upf.edu › activitats ›...
TRANSCRIPT
Barcelona listeners’ auditory
impressions from a perceptive
point of view within the
framework of Amper
Wendy Elvira (UB) Josefina Carrera (UB) Ana Ma. Fernández Planas (UB) Paolo Roseano (UB-UPF) Domingo Román (U. Católica de Chile) Eugenio Martínez Celdrán (UB)
Index • Introduction
• Methodology – The software
– Listeners
– The experiments
• Identification of modality – Results for interrogatives
– Results for declaratives
– Average
– Reaction time
• Cohesion between dialects – Results for declaratives
– Results for interrogatives
– Reaction time
• Conclusions
• References
Introduction
• Are Catalan speakers able to correctly
identify the modality of Romance
languages?
• How they establish if two dialects are the
same when there are big acoustic
differences?
Introduction
• Prosody-perception studies – Modality identification follows universal rules when it
is not biased by the mother tongue (Gussenhoven, 2000)
– It has been claim that language discrimination by prosody is possible even in early stages of life (Mehler et al. 1988). A little research has been made with close related languages, though (Vicenik, 2013).
• Acoustic studies – Amper-Cat and acoustic differences among Romanic
languages (Fernández Planas et al. 2011)
– Language contact and prosody borrowing (Martínez Celdrán, 2011; Romera y Elordieta, 2013)
Methodology
• Female voice re-synthesis
– Without lexical content
– Keeps linguistic rhythm and intonational features
– Only paroxitone configurations (SVO)
– Same number of syllables
• 11 stimuli par modality
– test 1: 22
– test 2 y 3: 11
Stimuli
Barcelona
València
L’Alguer
• Spanish
Palencia
La Laguna
Granada
• Sardinian
Biddanoa Monteleone
• Italian
Siena
Pòrtu Turre
• Furlan
Gardiscja
Beivars
• Catalan
The software
• Boersma & Weenik’s Praat
– Experiment MFC 6
• For identification and discrimination tests
– The experiments were concatenated into a
collection and worked in a bigger script.
– The script used recollected the listener’s data,
ran the experiments and save them in a
previously determined order.
Listeners
• 31 students (male and female) from the
University of Barcelona
• Bilingual (Catalan-Spanish)
• Sub-classified by their L1
– (this variable will only be taken into account if
differences between listeners exist)
• training in synthetic stimulus given before
the task
The experiments
• Forced choice tests
• Second audition possible
• Oops button
• Break every each 25 trials
• ISI 800ms
Identification of modality
• Instruction: Please, say whether the
sentence you hear is more likely to be a
question or a statement.
Test type: Identification
(top-down processing)
Number of stimuli: 22
Number of trials: 44
Presentation: individually
Results (identification of modality) Declaratives Interrogatives Average
Italian (Siena) 3% 8% 6%
Catalan (València) 8% 5% 6%
Spanish (Palencia) 7% 8% 7%
Catalan (Barcelona) 15% 2% 8%
Spanish (Granada) 15% 8% 11%
Friulian (Gardisca) 23% 8% 15%
Friulian (Beivars) 60% 16% 38%
Spanish (La Laguna) 42% 39% 40%
Catalan (L'Alguer) 26% 66% 46%
Italian (Pórtu Turre) 18% 77% 48%
Sardinian (Biddanoa) 7% 95% 51%
0%10%20%30%40%50%60%70%80%90%
100%
Italia
n (S
iena
)
Cat
alan
(Valèn
cia)
Spa
nish
(Palen
cia)
Cat
alan
(Bar
celona
)
Spa
nish
(Gra
nada
)
Furla
n (G
ardisc
a)
Furla
n (B
eiva
rs)
Spa
nish
(La
Lagu
na)
Cat
alan
(L'A
lgue
r)
Italia
n (P
ortu
Tur
re)
Sar
dini
an (B
idda
noa)
Declaratives
Interrogatives
Interrogatives
Spanish falling interrogatives
¡H* L%
A echo pattern in Catalan:
L+¡H* L%
Declaratives
Declaratives and interrogative
average
Cohesion between dialects
• Different dialects encode modality differently even when they belong to the same linguistic domain.
• So, do the listeners perceive the statements and questions of these dialects as if they were one? Or are they influenced by the acoustic cues and the would label them as different?
Discrimination test
• Do you think that this two sentences could belong to the same dialect?
Test type: AX (bottom-up processing)
Number of tests: 2
Interrogative yes/no sentences
Declarative sentences
Logical possibilities: AA/AA’ o AB
Number of trials: 121
Results declaratives
Descriptive stadistics
Multidimensional scaling
Cluster analysis: euclideic distance
Groups out of previous group.
BCN L'Alguer València Palència Granada La Laguna Gradisca Beivars Siena PortuTurre Biddanoa
BCN 93,5 45,2 80,6 41,9 22,6 9,7 58,1 45,2 38,7 35,5 19,4
L'Alguer 83,9 96,8 83,9 25,8 48,4 22,6 51,6 45,2 35,5 51,6 19,4
València 77,4 58,1 100 41,9 61,3 25,8 45,2 29 48,4 48,4 9,7
Palència 51,6 32,3 38,7 100 51,6 32,3 51,6 51,6 64,5 38,7 16,1
Granada 41,9 48,4 77,4 48,4 100 61,3 41,9 74,2 58,1 32,3 16,1
LaLaguna 32,3 32,3 41,9 48,4 54,8 100 54,8 51,6 48,4 35,5 12,9
Gradisca 51,6 38,7 51,6 48,4 38,7 51,6 96,8 87,1 51,6 38,7 12,9
Beivars 58,1 41,9 58,1 41,9 58,1 38,7 67,6 96,8 38,7 48,4 6,5
Siena 35,5 22,6 67,6 64,5 51,6 38,7 48,4 32,3 100 51,6 9,7
PortuTurre 58,1 38,7 48,4 32,3 29 32,3 45,2 35,5 58,1 100 22,6
Biddanoa 19,4 22,6 0 16,1 9,7 16,1 29 16,1 12,9 35,5 100
Results interrogatives
Descriptive stadistics
Multidimensional scaling
Cluster analysis: euclideic distance
Groups out of previous group.
BCN L'Alguer València Palència Granada LaLaguna Gradisca Beivars Siena Portu Turre Biddanoa
BCN 96,8 64,5 96,8 54,8 61,3 25,8 29 67,7 54,8 25,8 6,5
L'Alguer 51,6 96,8 67,7 35,5 41,9 22,6 32,3 45,2 32,4 16,1 19,4
València 93,5 61,3 100 32,3 48,4 22,6 51,6 77,4 45,2 29 6,5
Palència 54,8 45,2 41,9 100 45,2 48,4 32,3 54,8 35,5 19,4 12,9
Granada 90,3 45,2 71 64,5 100 35,3 45,2 71 48,4 22,6 12,9
LaLaguna 45,2 41,9 32,3 48,4 41,9 96,8 25,8 25,8 29 22,6 9,7
Gradisca 45,2 54,8 71 41,9 45,2 38,7 100 87,1 35,5 48,4 19,4
Beivars 67,7 45,2 67,7 38,7 71 32,3 74,2 96,8 45,2 35,5 19,4
Siena 61,3 45,2 51,6 58,1 51,6 25,8 32,3 48,4 100 41,9 12,9
PortuTurre 25,8 54,8 45,2 35,5 22,6 12,9 38,7 45,2 35,5 100 25,8
Biddanoa 12,9 25,8 12,9 12,9 3,2 3,2 19,4 9,7 9,7 29 93,5
Declaratives
Interrogatives
Conclusions
• Identification of modality
– cues to identify modality in both general and specific
patterns of their language.
– In patterns that are not alike but have a resemblance
• They find that a stimulus can be of a same
dialect attending to:
– its belonging or not to their mother tongue
– acoustic intonational resemblances
Gràcies!
References
•Gussenhoven, C., & Chen, A. (2000). Universal and language-specific effects in the
perception of question intonation. In Proceedings of the 6th ICSLP (pp. 91-94).
•Fernández Planas, A.M.; Roseano, P.; Dorta, J. i Martínez Celdrán, E. (forthcoming):
«¿Continuidad prosódica en diferentes puntos de la Romania? El caso de algunas
interrogativas», dins Actes del 26è Congrés Internacional de Lingüística i Filologia
Romàniques, València.
•Fernández Planas, A.M.; Roseano, P.; Martínez Celdrán, E. i Romera, L.
(2011):«Aproximación al análisis dialectométrico de la entonación en algunos puntos del
dominio lingüístico catalán», Estudios de Fonética Experimental, Vol. XX, pp. 141-178.
•Mehler, J; Jusczyk, P W.; Lambertz, G.; Hasltead, N; Bertoncini J; Amied-Tison, C.
(1988) A precursor of language acquisition in young infants. Cognition, 29, 143-178.
•Romera, Magdalena, and Gorka Elordieta. "Prosodic accommodation in language
contact: Spanish intonation in Majorca." International Journal of the Sociology of
Language 2013.221 (2013): 127-151.
•Vicenik, Chad, and Megha Sundara. "The role of intonation in language and dialect
discrimination by adults." Journal of Phonetics 41.5 (2013): 297-306.
Acknowledgements
• This project is feasible thanks to FFI2009-
09309/FILO and FFI2012-35998.
• The Laboratory of Phonetics (UB) research
group
• To the speakers
• To the listeners