Optimality-Theoretic modelling of phoneme split Paul Boersma, University of Amsterdam Paola Escudero, McGill University EuroSLA 11, Paderborn September.

Slides:

Advertisements

Similar presentations

Tom Lentz (slides Ivana Brasileiro)

Advertisements

Turning an L1 three-way contrast into an L2 two-way contrast Paola Escudero University of Utrecht and McGill University Paul Boersma University of Amsterdam.

Effects of Competence, Exposure, and Linguistic Backgrounds on Accurate Production & Perception of English Pure Vowels by Native Japanese and Mandarin.

Tone perception and production by Cantonese-speaking and English- speaking L2 learners of Mandarin Chinese Yen-Chen Hao Indiana University.

Human Speech Recognition Julia Hirschberg CS4706 (thanks to John-Paul Hosum for some slides)

1 CS 551/651: Structure of Spoken Language Spectrogram Reading: Stops John-Paul Hosom Fall 2010.

Copyright © 2009 Pearson Education Canada6-1 Chapter 6: Cognition in Infants and Toddlers 6.1 Piaget’s Theory 6.2 Information Processing 6.3 Language MODULES.

One-to-multiple vowel mapping in the perception of Dutch learners of Spanish Paola Escudero McGill University and University of Utrecht Paul Boersma University.

Psych 156A/ Ling 150: Acquisition of Language II Lecture 3 Sounds.

The Perception of Speech. Speech is for rapid communication Speech is composed of units of sound called phonemes –examples of phonemes: /ba/ in bat, /pa/

Speech perception 2 Perceptual organization of speech.

Development of Speech Perception. Issues in the development of speech perception Are the mechanisms peculiar to speech perception evident in young infants?

The Perception of Speech. Speech is for rapid communication Speech is composed of units of sound called phonemes –examples of phonemes: /ba/ in bat, /pa/

Modelling the perceptual development of phonological contrasts with OT and Gradual Learning Algorithm Paola Escudero, University of Reading

Second Language Acquisition: Introduction Paola Escudero Optimality Theory and Phonological Acquisition Seminar, Utrecht Institute of Linguistics OTS March.

Modelling the perceptual development of first- and second-language phonological contrasts with the Gradual Learning Algorithm Paola Escudero, University.

Exam 1 Monday, Tuesday, Wednesday next week WebCT testing centre Covers everything up to and including hearing (i.e. this lecture)

One-to-multiple vowel mapping in the perceptual development of Dutch learners of Spanish Paola Escudero McGill University and University of Utrecht Paul.

Differences in production account for differences in perception: The Gradual Learning Algorithm predicts language-specific cue weighting Paul Boersma,

Phonological constraints as filters in SLA Raung-fu Chung

Psych 56L/ Ling 51: Acquisition of Language Lecture 8 Phonological Development III.

SPEECH PERCEPTION The Speech Stimulus Perceiving Phonemes Top-Down Processing Is Speech Special?

Research on teaching and learning pronunciation

Probabilistic models in Phonology John Goldsmith University of Chicago Tromsø: CASTL August 2005.

Modelling the perceptual development of phonological contrasts with OT and the GLA Paola Escudero Paul Boersma

The Perception of Speech

NOVA Comprehensive Perspectives on Child Speech Development and Disorders Chapter 14 Acquisition of the English Voicing Contrast by Native Spanish-Speaking.

Phonemics LIN 3201.

SPEECH ARTICULATION: Vowels David Brett David Brett.

Segment Duration and Vowel Quality in German Lexical Stress Perception Klaus J. Kohler University of Kiel, Germany Paper presented at Speech Prosody 2012.

Saturday, March 15 th and Monday, March 17 th English FL: Reading Comprehension and Composition. Writing: Paragraph Structure; unity; parts, etc. Translation.

[kmpjuteynl] [fownldi]

Sebastián-Gallés, N. & Bosch, L. (2009) Developmental shift in the discrimination of vowel contrasts in bilingual infants: is the distributional account.

Speech Perception 4/6/00 Acoustic-Perceptual Invariance in Speech Perceptual Constancy or Perceptual Invariance: –Perpetual constancy is necessary, however,

Psych 156A/ Ling 150: Psychology of Language Learning Lecture 5 Sounds III.

Adaptive Design of Speech Sound Systems Randy Diehl In collaboration with Bjőrn Lindblom, Carl Creeger, Lori Holt, and Andrew Lotto.

C. Language Transfer Odlin, 1989; Hayes, 2003; Wu, 1987 Positive transfer : learners use the patterns or the sounds of their first language to help themselves.

Acoustic Cues to Laryngeal Contrasts in Hindi Susan Jackson and Stephen Winters University of Calgary Acoustics Week in Canada October 14,

Demonstration and Verbal Instructions

Intelligibility of voiced and voiceless consonants produced by Lebanese Arabic speakers with respect to vowel length Romy Ghanem.

Assessment of Phonology

Epenthetic vowels in Japanese: a perceptual illusion? Emmanual Dupoux, et al (1999) By Carl O’Toole.

© 2008 The McGraw-Hill Companies, Inc. Chapter 8: Cognition and Language.

3308 First Language acquisition Acquisition of sounds Perception Sook Whan Cho Fall, 2012.

Chapter 2 Chapter 2 Teaching Pronunciation. I why teach pronunciation? 1. Inaccurate production of a phoneme or inaccurate use of suprasegmental elements.

A Study of Taiwanese High School Students' Production and Perception Performance in English Non-High Front Vowels Graduate Student: Wan-chun Tseng Advisor:

The Holes in the Brian Help Us Sort Out Sounds..  I. The Brain’s ability to sort out sounds  1. speech sounds are categorized.  2.Misinterpretations.

1 Lesson 4 Attitudes. 2 Lesson Outline   Last class, the self and its presentation  What are attitudes?  Where do attitudes come from  How are they.

Na1c0014 李羿霈.  An acoustic perspective of English vowel production and perception by Taiwanese EFL learners, as compared with native speakers of English.

Acoustic Continua and Phonetic Categories Frequency - Tones.

1 Cross-language evidence for three factors in speech perception Sandra Anacleto uOttawa.

CSD 2230 INTRODUCTION TO HUMAN COMMUNICATION DISORDERS Normal Sound Perception, Speech Perception, and Auditory Characteristics at the Boundaries of the.

Neurophysiologic correlates of cross-language phonetic perception LING 7912 Professor Nina Kazanina.

Source of change –Combination of feedback and explain- experimenter’s-reasoning led to greater learning than feedback alone Path of change –Children relied.

The phonology of Hakka zero- initials Raung-fu Chung Southern Taiwan University 2011, 05, 29, Cheng Da.

RQ 2 No significant difference.. Results and Discussion RQ 3: Are the abilities of distinguishing among the two pairs of vowels in perception and production.

The Acquisition of English Vowels Reflected in Taiwanese EFL Learners Raung-fu Chung Southern Taiwan University April 27, Hechun.

Based on Lai Yi-shiu (2009). Cognitive linguistics.

0 / 27 John-Paul Hosom 1 Alexander Kain Brian O. Bush Towards the Recovery of Targets from Coarticulated Speech for Automatic Speech Recognition Center.

Training Spanish vowels Carmen Lie-Lahuerta Research Institute ACLC University of Amsterdam CAPSLaP 2014.

WebCT You will find a link to WebCT under the “Current Students” heading on It is your responsibility to know how to work WebCT!

Psych 156A/ Ling 150: Psychology of Language Learning Lecture 3 Sounds I.

Investigating the basis for conversation between human and robot Experiments using natural, spontaneous speech, speaking to the robot as if it were a small.

Bridging the gap between L2 speech perception research and phonological theory Paola Escudero & Paul Boersma (March 2002) Presented by Paola Escudero.

Chinese Learners’ Perception and Production of the vowels: /e/, /ei/, /o/, & /ou/ in English by Contrastive Analysis 研究生 : 張悅寧報告人 : NA2C0003 傅學琳 WHO WHAT.

The Prominent Errors in Vowel Sounds Committed by Lower Secondary Schools in Malaysia By: NURUL FARAH NADIA BINTI BAHARUM P84388.

Teaching pronunciation

Theoretical Discussion on the

How to Be a Successful Language Learner

Fundamentals of Sensation and Perception

Presentation transcript:

Optimality-Theoretic modelling of phoneme split Paul Boersma, University of Amsterdam Paola Escudero, McGill University EuroSLA 11, Paderborn September 26, 2001

Six hypotheses about phonemic contrasts ¿ Their production is dialect-dependent. ¿ Their perception is dialect-dependent. ¿ L1 learners achieve optimal perception given their production environment. ¿ This can be modelled by Optimality Theory and the Gradual Learning Algorithm. ¿ L2 learners do not necessarily achieve optimal perception. ¿ This can also be modelled by OT & GLA.

We test these six hypotheses on the English /I/-/i/ contrast L1: Scottish Standard English L1: Southern English Standard English L2: Spanish-speaking learners of English

Adult native Scottish production

Adult native Southern production

Six hypotheses about phonemic contrasts  Their production is dialect-dependent. ¿ Their perception is dialect-dependent. ¿ L1 learners achieve optimal perception. ¿ This can be modelled by OT & GLA. ¿ L2 learners do not necessarily achieve optimal perception. ¿ This can also be modelled by OT & GLA.

Adult native perception Scottish Southern

Six hypotheses about phonemic contrasts  Their production is dialect-dependent.  Their perception is dialect-dependent. ¿ L1 learners achieve optimal perception. ¿ This can be modelled by OT & GLA. ¿ L2 learners do not necessarily achieve optimal perception. ¿ This can also be modelled by OT & GLA.

Why does perception have to depend on the production environment?  Answer: to optimise perception, the listener has to minimise the probability of perceptual confusion  The optimal perception strategy, therefore, is: likelihood maximisation, i.e. choose the most likely produced category, given a certain F1 & duration

Average production environment: Optimal perception (max. likelihood): 87.1% 82.5% Scottish Southern

Six hypotheses about phonemic contrasts  Their production is dialect-dependent.  Their perception is dialect-dependent.  L1 learners achieve optimal perception (they integrate the two acoustic cues). ¿ This can be modelled by OT & GLA. ¿ L2 learners do not necessarily achieve optimal perception. ¿ This can also be modelled by OT & GLA.

Native English constraint set “260 Hz should not be perceived as /I/” “260 Hz should not be perceived as /i/” “500 Hz should not be perceived as /I/” “500 Hz should not be perceived as /i/” “60 ms should not be perceived as /I/” “60 ms should not be perceived as /i/” “180 ms should not be perceived as /I/” “180 ms should not be perceived as /i/” …and so on, for all F1 and duration values

Average production environment: Optimal perception (max. likelihood): 87.1% 82.5% Scottish Southern

Scottish optimal perception in OT [349 Hz, 74 ms] 349 Hz not / I / 74 ms not / i / 74 ms not / I / 349 Hz not / i / /I/ /I/ *!* /i//i/ ** [349 Hz, 74 ms] 349 Hz not / i / 74 ms not / i / 74 ms not / I / 349 Hz not / I / /I//I/ ** /i/ /i/ *!* Southern optimal perception in OT

How is the knowledge acquired?  Whenever the listener makes a categorization error, she applies a Gradual Learning Algorithm: [349 Hz, 74 ms] 349 Hz not / i / 74 ms not / i / 74 ms not / I / 349 Hz not / I / /I//I/ ** ** /i//i/ *!  **

Simulated Scottish L1 development baby 4 months 1000 months 50.0% 85.3% 87.0% (optimal: 87.1%)

Simulated Southern L1 development baby 4 months 1000 months 50.0% 79.7% 82.5% (optimal: 82.5%)

Six hypotheses about phonemic contrasts  Their production is dialect-dependent.  Their perception is dialect-dependent.  L1 learners achieve optimal perception.  This can be modelled by OT & GLA. ¿ L2 learners do not necessarily achieve optimal perception. ¿ This can also be modelled by OT & GLA.

Typical L2 categorization by Spanish learners of English ScotlandSouthern England

Six hypotheses about phonemic contrasts  Their production is dialect-dependent.  Their perception is dialect-dependent.  L1 learners achieve optimal perception.  This can be modelled by OT & GLA.  L2 learners do not necessarily achieve optimal perception. ¿ This can also be modelled by OT & GLA.

Native Spanish constraint set “260 Hz should not be perceived as /i/” “260 Hz should not be perceived as /e/” “260 Hz should not be perceived as /a/” “260 Hz should not be perceived as /o/” “260 Hz should not be perceived as /u/” low ranked or non-existent: “60 ms should not be perceived as /i/” “60 ms should not be perceived as /e/” “60 ms should not be perceived as /a/” “60 ms should not be perceived as /o/” “60 ms should not be perceived as /u/” …and so on, for all F1 and duration values

Constraint set of Spanish learners of English “260 Hz should not be perceived as /i/” “260 Hz should not be perceived as /e/” “260 Hz should not be perceived as /a/” “260 Hz should not be perceived as /o/” “260 Hz should not be perceived as /u/” “60 ms should not be perceived as /short/” “60 ms should not be perceived as /long/” …and so on, for all F1 and duration values, but: no F1-to-/length/ mappings; no duration-to-/vowel quality/ mappings.

Strategy of a Spaniard in Scotland Two-category assimilation: English /i/  L2 /i/ English / I /  L2 /e/

Simulation of a Spaniard in Scotland 200 months 204 months 1000 months 74.9% 83.5% 87.2% (optimal: 87.1%)

Strategy of a Spaniard in S.England New length contrast: English /i/  L2 /i:/ English / I /  L2 /i/

Simulation of a Spaniard in S.England 200 months 204 months 1000 months 50.0% 79.0% 79.3% (optimal: 82.5%)

Six hypotheses, now finally confirmed by us, about phonemic contrasts  Their production is dialect-dependent.  Their perception is dialect-dependent.  L1 learners achieve optimal perception.  This can be modelled by OT & GLA.  L2 learners do not necessarily achieve optimal perception.  This can also be modelled by OT & GLA.

Native S.English has 12 arbitrary symbols: A possible assimilation pattern Spanish S.English has 2x5 vowels (7 symbols): I — i i — ii — i ,  — A E — e  — o U — u e — ee — e a — aa — a o — oo — o u — uu — u

Discussion  GLA generally leads to optimal perception, but it depends on the appropriateness of the constraint set.  L2 learners seem to have constraint sets that are too restricted for the new language environment.  The restricted constraint set does not seem to cause the L2 learner much harm.  Adding a new contrast on the basis of duration seems to be easier than splitting existing vowel quality categories.

Thank you for your attention!