Session Poster V:

Poster V

Type: poster
Chair: Petra Wagner, Esther Janse
Date: Thursday - August 09, 2007
Time: 10:00
Room: Poster Area


Poster V-1 When does lip protrusion start in Standard Austrian German? An acoustic investigation
Sylvia Moosmüller, Acoustics Research Institute of the Austrian Academy of Sciences
Paper File
  CV, V1CV2- and V1#CV2-sequences of reading material of six speakers of Standard Austrian German have been analysed. V1 was a pre-palatal, constricted vowel /i/ in unstressed position, C an alveolar consonant, and V2 either a pre-palatal, constricted vowel /i/ or a back, rounded vowel /u, o/ in a stressed position. F1, F2, F3, and VOT measurements were performed. Lip protrusion starts at consonant release and may affect the transconsonantal vowel, as long as V1 and C are not separated by a word boundary.
Poster V-3 A Comparison of Vowel Acoustics Between Older and Younger Adults
Peter Watson, University of Minnesota - Twin Cities
Benjamin Munson, University of Minnesota - Twin Cities
Paper File
  Previous research has shown a difference in vowel acoustics between older and younger adults, possibly related to age-related changes in vocal tract morphology. Other data suggest that vowel acoustics may vary as a function of neighborhood density and word frequency in older adults, possibly due to the mediating influence of lexical access. This investigation examined whether these two factors interact. Results show that older adults had overall lower-frequency formants, and qualitatively different-shaped vowel spaces, than the younger adults, but the influences of word frequency and neighborhood density on the acoustic characteristics of vowels were statistically equivalent in both groups.
Poster V-5 Lingual Contact in selected English Vowels and its acoustic consequence
Ivan Yuen, Department of Psychology, Royal Holloway, University of London
Alice Lee, Department of Speech and Hearing Sciences, University College Cork
Fiona Gibbon, Department of Speech and Hearing Sciences, Queen Margaret University College
Paper File
  This paper provides preliminary data about EPG contact for 3 different vowels in Southern British English and Scottish English across eleven speakers. The EPG data were compared with vowel formants to test the hypothesis that the amount of EPG contact as an indicator of tongue height or anteriority will result in a corresponding change in F1 and F2. The results suggest that Percent Contact varies with the three monophthongs. F1, F2 and F2-F1 difference varies with the amount of Percent Contact.
René Carré, Laboratoire Dynamique Du Langage (DDL), CNRS-Université de Lyon 2
François Pellegrino, Laboratoire Dynamique Du Langage (DDL), CNRS-Université de Lyon 2
Pierre Divenyi, VA Medical Center and EBIRE, Martinez, California, USA
Paper File
  Speech is generally looked upon as a succession of events in the time domain and analyzed frame by frame, while ignoring the fact that speech is dynamic. In the present paper, evidence in support of the dynamic nature of speech and dynamic invariance, as well as their consequences on speech research, are discussed.
Poster V-9 The comprehension of acoustically reduced morphologically complex words: The roles of deletion, duration, and frequency of occurrence
Mirjam Ernestus, Radboud University Nijmegen & Max Planck Institute for Psycholinguistics
Harald Baayen, Radboud University Nijmegen & Max Planck Institute for Psycholinguistics
Paper File
  This study addresses the roles of segment deletion, durational reduction, and frequency of use in the comprehension of morphologically complex words. We report two auditory lexical decision experiments with reduced and unreduced prefixed Dutch words. We found that at the macro level, segment deletions lead to delayed comprehension. At the micro level, however, longer durations appear to increase lexical competition, either from the word's stem (Experiment 1) or from the word's morphological continuation forms (Experiment 2). Increased lexical competition slows down especially the comprehension of low frequency words, which shows that speakers do not try to meet listeners' needs when they reduce especially high frequency words.
Poster V-11 Interlingual near homophonic words and phrases in L2 listening: Evidence from misheard song lyrics
Takashi Otake, E-Listening Laboratory
Paper File
  Recent studies on bilingual spoken-word recognition have shown that bilinguals cannot deactivate the lexicon of the native language in a monolingual non-native situation due to the fact that language nonselective access is applied. This study attempted to examine whether this phenomenon could be observed in a unique Japanese word play called soramimi awaa in which Japanese misheard song lyrics were extracted from foreign songs. The analysis showed that both interlingual near homophones and phrases were observed, suggesting that at least Japanese bilinguals may experience the deactivation of the lexicon in their linguistic activity.
Baris Kabak, University of Konstanz
Kazumi Maniwa, University of Konstanz
Paper File
  This study investigated perception by non-native listeners of English fricatives produced in clear and conversational speaking styles. We measured babble thresholds for fricative voicing and place of articulation contrasts by Standard German and Swabian German and native American English speakers. Overall, Swabian German speakers performed worse than both native English and Standard German speakers, and Standard German speakers worse than native English speakers. German speakers in general had more difficulty with non-sibilant distinctions, and Swabian speakers also had difficulty with sibilant voicing distinctions. A robust clear speech benefit was observed across groups and contrasts. Overall, the results indicate that difficulty in perceiving foreign-language contrasts stems from the interaction of phonological, phonetic, and psychophysical issues.
Poster V-15 Perceptual Effect of Vowel Devoicing and Its Working Range
Makiko AOYAGI, Dokkyo University
Paper File
  This study reports a perceptual effect of a devoiced vowel in Japanese as an assimilative outcome of connected speech. A focus is placed on how such an effect changes its magnitude as the source and the recipient of the effect are separated. A devoiced syllable ki- has a strong effect of causing voiceless judgment in the following da-ta VOT variations with natural closure durations. However, such an effect is weakened as the closure interval is expanded beyond the original one for a voiceless stop. In this case, the original voicing judgment of the da-ta variations in isolation gradually returns. Also, the more voiced the da-ta variation itself, the earlier and farther it deviates from the influence of the devoiced vowel. Phonetic variation resulting from connected speech serves to aid segmental perception, but in turn the effect obtains in a ‘connected’ speech event.
Poster V-17 The acoustic analysis of Persian fricative-affricate contrast
Zahra Mahmoodzadeh, Department of Linguistics, University of Tehran
Mahmoud Bijankhan, Department of Linguistics, University of Tehran
Paper File
  Acoustic analysis suggests that each of the following variables can cue the post-alveolar fricative-affricate contrast in isolated word forms in Persian: silence duration, frication duration, rise time and amplitude rise slope. The acoustic values of each cue differ with the position of the test item in the word. Silence and frication duration of Persian affricates was longest in the final position. The rise time of Persian voiceless affricate was also longest finally but the amplitude rise slope was longest initially. It is confirmed that there is a positive relation between frication duration and silence duration, frication duration and rise time and also silence duration and rise time. The trading relation between frication duration and amplitude rise slope was negative.
Poster V-19 Robustness of acoustic landmarks in spontaneously-spoken American English
Stefanie Shattuck-Hufnagel, Speech Group, RLE, Massachusetts Institute of Technology
Nanette Veilleux, Speech Group, RLE, Massachusetts Institute of Technology, and Simmons College
Paper File
  Acoustic landmarks (abrupt changes associated with consonant closures and releases, vowels and glides) play an important role in some models of lexical access (e.g. Stevens 1998, 2002), so it is important to determine how often they survive articulatory overlap and weakening in spontaneous speech production. A corpus of spontaneous American English speech collected from 8 adult female speakers is hand labeled for the occurrence of landmarks. Preliminary results for one conversation (240 secs., 610 words, analysis completed for 1003 of 2750 predicted landmarks) show that 86% of landmarks were realized overall, with a sharply lower rate for coronal stops /t/ and /d/. These results suggest that the majority of landmarks are available for detection both by human listeners and automatic recognition algorithms. Ongoing analyses are comparing the rate of automatic detection of these acoustic events with the hand labels, and tabulating the contexts in which landmarks are lost or changed.
Paper File
  This investigation is part of a larger study of the role of fine phonetic details in word segmentation in Greek connected speech. The present paper investigates whether and how Greek speakers use durational and pitch alignment acoustic cues to mark word boundaries in identical segmental strings differing only in the word boundary affiliation. Duration modification mechanisms are evident in cuing words, while different F0 alignment is not detected.
Poster V-23 The Phonetics-Phonology Interface of Erzya Stress: Morphological Conditioning of Vowel Reduction
NIINA AASMÄE, University of Tartu
JAAN ROSS, University of Tartu
Paper File
  Analyses provided in this paper test the possibility of morphological conditioning for vowel durations in Erzya. The duration of stressed first-syllable and unstressed second-syllable vowels in disyllabic ‘root’ and ‘inflectional form’ tokens was compared. Material consisted of one-word utterances produced in spontaneous speech by speakers of different dialects. The inter-idiolect vowel durations tended to equalize within the root. At the boundary of the root and a suffix vowels were shorter than in the stressed first syllable. The manifestations of the tendency varied across four idiolect groups representing the main language varieties. The results suggest that there is a causal relationship between unstressed vowel reduction and the domain of morphology. Dialects revealing variability in the manifestation of the general tendency towards vowel duration asymmetry in the root and inflectional forms also display differences in the mobility of stress, as suggested in previous research.
Kikuo Maekawa, National Institute for Japanese Language
Yosuke Igarashi, National Institute for Japanese Language
Paper File
  Lexical pitch accents in bimoraic particles of Tokyo Japanese are believed to be deleted when the particles are combined with accented words. Analysis of the Corpus of Spontaneous Japanese revealed, however, there are many cases where particles retain their accent, thereby forming an accentual phrase of their own. Factors that favor accent preservation include semantic properties of particles, inter-accent distance, boundary pitch movements, and, formality of speech. Particles having emphatic and/or limitative meanings, like /sa’e/, /ko’so/ and /no’mi/, are the most probable to retain their accents.
Johanneke Caspers, Dept. of Dutch Studies / Leiden University Centre for Linguistics, Leiden University
Paper File
  A distinction can be made between three different final boundary tone types in Dutch: high (H%), low (L%) and ‘level’ (%). As yet it is not completely clear what these tones signify to the listener; the present perception test aims to find out more about the interpretation of these tones. In a listening experiment declarative sentences with different intonation contours were presented to native listeners, who had to indicate whether they thought the stimulus ended in a comma, a full stop or a question mark. Results are clear for L% tones (full stop) and for level tones (mainly commas, no question marks). Within the category of H% tones the higher ending rises lead to more question responses, while the lower ending rises prompt more comma responses.
Poster V-29 Prosodic Boundary Effects on Durations and Vowel Hiatus in Modern Greek
Evia Kainada, University of Edinburgh
Paper File
  Research on the identification of the prosodic structure of languages has been based on phonetic processes, such as durational patterns and sandhi phenomena. One of the main assumptions is that such processes all signal, and are thus regulated by, the same structure. The experiment reported here tests the validity of this assumption by investigating whether prosodic boundary strength has the same effect on various segmental processes. The application of pre- and post-boundary duration and of vowel hiatus is investigated under different prosodic conditions in Modern Greek. Preliminary results suggest that there is a tendency for a similar effect of boundary strength on both processes, with vowel hiatus showing potentially a different application in one of the conditions.
Poster V-31 Map Task Dialogs in Noise - a Paradigm for Examining Lombard Speech
Hansjörg Mixdorff, Department of Computer Sciences and Media, Berlin University of Applied Sciences
Ulrich Pech, Department of Computer Sciences and Media, Berlin University of Applied Sciences
Chris Davis, MARCS Auditory Laboratories, University of Western Sydney
Jeesun Kim, MARCS Auditory Laboratories, University of Western Sydney
Paper File
  This paper presents a paradigm for comparing auditory-visual map task dialogs produced in silence and in noise, also known as Lombard speech. A previously developed temporal filtering algorithm which removes the ambient noise from recordings of Lombard speech was modified to accommodate longer recordings. On a small production dataset of two levels of vehicle and babble noise we examined the effect on fundamental frequency and intensity contours. We found that Lombard characteristics of speech, that is, an increase in mean F0 as well as intensity, are stronger for babble than for vehicle noise. There are indications that talkers become habituated to the noisy environment during a dialog. Participants appeared to solve the task more leisurely in silence than in noise. By performing eye-tracking on one of the talkers' data we found that the frequency of gaze was more than double in babble noise than in silence.
Poster V-33 Single H and Doubly-linked H in South Kyungsang Korean
Seung-Eun Chang, University of Texas at Austin
Paper File
  This experiment examines the distinction between single H and doubly-linked H of monomorphemic words (HL, HH) and bimorphemic words (H+L, H+H) in South Kyungsang Korean. The results showed that F0 fall comes later in double H than in single H, and thus peak plateau is longer in double H than in single H. However, F0 timing difference was also found within double H, depending on the morpheme type, i.e., the peak plateau is longer in monomorphemic HH than in bimorphemic H+H. This suggests that the morphological structure may influence the phonetic realization such as F0 timing. In addition, this instrumental data do not confirm the H tone spreading analysis in suffixed words, not as suggested in earlier transcription based studies.
Poster V-35 Prosodic phrasing in tonal and non-tonal dialects of Kammu
Anastasia Mukhanova Karlsson, Lund University
David House, Kungliga Tekniska Högskolan
Jan-Olof Svantesson, Lund University
Damrong Tayanin, Lund University
Paper File
  Kammu is a language that has developed lexical tones rather recently. One dialect is a tone language with (high or low) tone on each syllable, while the other main dialect lacks lexical tones. The dialects differ only marginally in other respects. This allows us to investigate how the existence of lexical tones in a given language influences the use of intonation. We performed a study of tonal means of phrasing in tonal vs. non-tonal dialects. We did find differences in boundary signaling. In both types of dialects the differentiation between marked and unmarked boundaries is relevant. At marked phrase boundaries we find signaling of focus and of some expressive meanings. The difference between the dialects is in the functional load of the intonational gestures. In the tone dialects pragmatically marked boundaries are assigned high pitch, while in non-tonal dialect the falling tone has a high pragmatic load.
Paola Escudero, University of Amsterdam
Ricardo Augusto Hoffmann Bion, University of Amsterdam
Paper File
  This study constitutes the first attempt at combining vowel normalization procedures with the linguistic perception framework of Stochastic Optimality Theory [1] and the Gradual Learning Algorithm [2]. Virtual learners possessing different normalization procedures, and a control learner with no normalization, were trained to perceive Brazilian Portuguese and American English vowels. Our results show that learners equipped with normalization algorithms outperformed the control learners, obtaining accuracy scores up to 46% higher. Thus, this model in which normalization and sound perception are implemented as two sequential processes delivers the expected results. That is, it improves the performance of a perception grammar when the training and testing sets have speakers with different ages and gender.
Poster V-39 Directionality of tone change
Pittayawat Pittayaporn, Department of Linguistics, Cornell University
Paper File
  In this paper, I present a theory of tonal change focusing on the directionality of tone change. Drawing on studies on phonetic variation of tones, I propose three main mechanisms that govern the directions in which tones change: 1) segment-tone interaction, 2) contextual variation, and 4) perceptual maximization. The predictions made by a model that includes these mechanisms are borne out by the tonal development of Bangkok Thai.
Poster V-41 Impact of duration and vowel inventory size on formant values of oral vowels: an automated formant analysis from eight languages
Cédric Gendrot, L.P.P.
Martine Adda-Decker, L.I.M.S.I.
Paper File Additional Files
  Eight languages (Arabic, English, French, German, Italian, Mandarin Chinese, Portuguese, Spanish) with 6 differently sized vowel inventories were analysed in terms of vowel formants. A tendency to phonetic reduction for vowels of short duration clearly emerges for all languages. The data did not provide evidence for an effect of inventory size on the global acoustic space and only the acoustic stability of quantal vowel /i/ is greater than other vowels’ in many cases.
Poster V-43 A cross-dialect comparison of Peninsula- and Peruvian-Spanish vowels
Geoffrey Stewart Morrison, Department of Cognitive & Neural Systems, Boston University
Paola Escudero, Institute of Phonetic Sciences, University of Amsterdam
Paper File
  A comparison was made of the acoustic properties of Spanish vowels produced by monolingual Spanish speakers from Spain and Peru. Monophthongs were produced in sentence final position. Peninsula speakers’ vowels were shorter, had lower fundamental frequency, and were more likely to be produced with creaky voice. A multivariate test on the whole vowel system did not find a significant cross-dialect difference in formant values. Implications for second-language speech perception and production research are discussed.
Poster V-45 Investigating British Asian accents: Studies from Glasgow
Kirsten Lambert, Speech and Hearing Sciences, Queen Margaret University
Farhana Alam, Department of English Language, University of Glasgow
Jane Stuart-Smith, Department of English Language
Paper File
  Despite the substantial Asian community in the UK, there has been very little phonetic work on British Asian accents. The complementary results from two small-scale studies of Glasgow Asian accent, confirm the identification of Glasgow Asian as an recognizable accent, identify accent features particular to Glasgow Asian and not found in Glaswegian more generally, and confirm their use – with specific social-indexical functions – in everyday speech.
Yi-Hsuan Huang, Graduate Institute of Linguistics, National Taiwan University
Janice Fon, Graduate Institute of Linguistics, National Taiwan University
Paper File
  This paper aims to explore the effect of acquisition order and word-relatedness on code-switching costs in bilingual speakers. 38 Mandarin-Min bilinguals performed a picture-naming task, in which hand-drawn pictures were color-coded for the two languages, Mandarin and Min, and switching points were pre-determined but variable. Results showed that naming latencies of cognates were in general shorter than non-cognates, and Mandarin stimuli were also shorter than Min. Min non-cognates were especially difficult for subjects. Code-switched trials incurred longer latencies in subjects, but only in those who acquired both languages at the same time, contrary to what was predicted by the Inhibitory Control Model.
Masaki Taniguchi, Kochi University
Yusuke Shibata, Kochi University
Paper File Additional Files
  This research attempts to clarify the difference between Japanese learners’ intended tonicity and performed tonicity, i.e., between their knowledge and practice. The results were as follows: (1) The subjects tended to put a nucleus on the stressed syllable of the last word in each intonation phrase. They typically used high level pitch to highlight the word that they thought they had to put a nucleus on. (2) Their intended tonicity was strikingly better than their performed tonicity. (3) There was greater discrepancy between intended tonicity and performed tonicity when they had to find correct tonicity on their own than when they were provided with it. (4) We need to consider two kinds of errors: (a) error in intended tonicity (error in knowledge) and (b) error in performed tonicity (error in putting knowledge into practice). (5) Teaching tonicity requires providing with knowledge plus exercise to put knowledge into practice.
Bo Thorén, Dept of Linguistics, Stockholm University
Paper File
  In order to test the persistency of the Swedish complementary durational pattern of VC-sequences in stressed syllables, a number of native Swedish speakers were recorded when pronouncing words in English and German. The words were of a kind that were expected to be perceived by Swedes as having “short vowel”. Swedish speakers pronounced the test words with significantly longer post-vocalic stop consonant /k/ and /t/, than did native English and German speakers, but not when the test word contained a post-vocalic nasal /m/. This asymmetry was not found when native Swedish speakers pronounced Swedish words with the same segments in the VC-sequence. Keywords: foreign accent, temporal patterns, complementary consonant duration.
Poster V-53 Effects of Phonetic similarity and L2 experience: Production of English /s-sh/ by adult Korean ESL learners
Sang Yee Cheon, University of Hawaii at Manoa
Paper File
  This study examined the effects of phonetic similarity between L1 and L2 sounds and L2 experience on the production improvement of L2 English /s/ and // by adult Korean ESL learners. Two production judgment methods were employed. The result shows that a dissimilar L2 sound /s-sh/ was better produced than a similar L2 sound /s/ in terms of listener judgment. Adult L2 learners did not produce similar L2 sounds accurately, even with extensive L2 experience. However, in terms of acoustic-based judgment, ESL learners produced L2 sounds accurately regardless of proficiency level, resulting in a disparity in production performance between two production judgment methods. In assessing the production ability of L2 learners, more than one phonetic-based judgment in addition to listener judgment need to be taken into consideration.
Poster V-55 Comparison of Pitch Range in Finnish (L1) and Russian (L2)
Riikka Ullakonoja, University of Jyväskylä
Paper File
  The aim of the present study is to investigate, whether the pitch range of a speaker can vary according to the language he speaks. The hypothesis is tested on Finnish university students studying Russian as a second language before and after their stay in Russia. The global pitch range (max – min) is determined as well as the pitch range in different types of utterances (declarative, question, exclamation) also by superimposing pitch contours. It was discovered that the L2 learners have a narrower pitch range in L1 and L2 and less variable pitch. However, the results suggest that L2 experience seems to help most students to produce more native like pitch range, especially in questions.
Poster V-57 Analysis and Synthesis of Speaker Age
Susanne Schötz, Linguistics and Phonetics, Centre for Languages and Literature, Lund University
Paper File
  Speaker age is an important speaker-specific quality, which was investigated in the two studies presented here. The first study automatically extracted 161 acoustic features from six words produced by 527 speakers, and used normalised mean values to compare the features. Segment duration and sound pressure level (SPL) range were identified as two important acoustic correlates of age. The second study developed a research tool for analysis of speaker age by data-driven formant synthesis and age-weigthed linear interpolation to simulate an age between the ages of any two of four female differently-aged reference speakers. Evaluation of the tool revealed that speaker age may in fact be simulated using formant synthesis. Both studies will be used in further attempts to model and simulate speaker age.
Poster V-59 Intersegmental cohesion and syllable division in Polish
Pier Marco Bertinetto, Scuola Normale Superiore, Pisa
Sylwia Scheuer, Adam Mickiewicz University, Poznań
Katarzyna Dziubalska-Kołaczyk, Adam Mickiewicz University, Poznań
Maddalena Agonigi, Scuola Normale Superiore, Pisa
Paper File
  An experiment with Polish participants was devised in order to shed light on ‘intersegmental cohesion hierarchy’, with special regard to CC sequences. This hierarchy regulates the strength of the segments’ mutual attraction, obeying both universal and language-specific tendencies. The results show that Polish speakers, as contrasted to Italian ones, exhibit a finer cohesion scale due to the richer phonotactics to which they are attuned. In the approach advocated by the authors, syllabic structure is assumed to epiphenomenically emerge from the given hierarchy.
Poster V-61 Autism and lexical context effects on speech perception
Mitsuhiko Ota, University of Edinburgh
Mary Stewart, Heriot-Watt University
Paper File
  The view that weak central coherence in processing causes autism implies that autistic individuals should exhibit attenuated lexical context effects on speech perception. To test this hypothesis, we examined the degree to which phonetic categorization shifts to make the percept a known word (i.e., the 'Ganong effect') in a neurotypical population with varying degrees of autistic traits. Fifty-eight university students were given the Autism-Spectrum Quotient (AQ) and a segment identification test using two word-to-nonword VOT continua (kiss-giss and gift-kift). A significant negative correlation was found between the total AQ score and the identification shift that occurred between the continua. The AQ score did not correlate with scores on separately administered VOT discrimination, auditory lexical decision, or verbal IQ, ruling out enhanced auditory sensitivity, slower lexical access or higher intelligence as explanations of the AQ-related shift in phonetic categorization.
Mariam Hartinger, Queen Margaret University, Edinburgh
William Hardcastle, Queen Margaret University, Edinburgh
Fiona Gardiner, NHS Highlands
Paper File
  The paper presents preliminary results of a speech motor control study in hypokinetic dysarthria in Parkinson’s disease (PD). By means of EPG, the tongue contacts of two speakers with PD and two control speakers during the production of target words containing the initial and final stops /t/ were analysed in normal and loud condition as well as in complex sentences. The preliminary results showed no effects of increasing loudness on duration and on the number of tongue contacts in speakers with PD. Furthermore, frication of the stop /p/ to [f] was found for one speaker in the acoustic analysis.
Poster V-65 Sonorant segment quality in Russian emotional speech
Veronika Makarova, University of Saskatchewan
Valery Petrushin , 4i Consulting Group, Inc
Paper File
  The paper reports characteristics of sonorant segments (vowels and sonorant consonants) in Russian emotional speech. The authors describe the effects of segmental duration, energy, formants and dynamic ranges on the expression of emotion in Russian. The data come from RUSLANA, a database containing samples of neutral utterances and utterances with simulated emotions of surprise, happiness, anger, sadness and fear.
Poster V-67 Cross-modal perception of emotional speech
Pashiera Barkhuysen, University of Tilburg
Emiel Krahmer, University of Tilburg
Marc Swerts, University of Tilburg
Paper File
  We report on a perception experiment in which Czech participants rate the perceived emotional state of Dutch speakers. These speakers could either display a positive or a negative emotion, which was either real or acted. The Czech participants were confronted with these utterances, which they could not understand, in a bimodal (audiovisual) or a unimodal (audio or vision only) condition. It was found that acted emotional speech leads to significantly more extreme perceived emotion scores than non-acted emotional speech, where the difference between acted and real emotional speech is stronger for the negative than for the positive conditions. Interestingly, the biggest overall differences between acted and non-acted emotions were found for the audio-only condition, which suggests that acting has a particularly strong effect on the spoken realization of emotions.
Bianca Vieru-Dimulescu, LIMSI-CNRS
Philippe Boula de Mareüil, LIMSI-CNRS
Martine Adda-Decker, LIMSI-CNRS
Paper File
  The goal of this study is to investigate the most relevant cues that differentiate foreign accents in French (Arabic, English, German, Italian, Spanish and Portuguese). We took advantage of automatic alignment into phonemes of non-native French recordings. Starting from standard acoustic models, we introduced pronunciation variants which were reminiscent of foreign-accented speech: first allowing alternations between French phonemes (e.g. [s]~[z]), then combining them with foreign acoustic units (e.g. a rolled r). Results reveal discriminating accent-specific pronunciations which, to a large extent, confirm both linguistic predictions and human listeners’ judgments.

Back to Conference Schedule