Uitspraakevaluatie & training met behulp van spraaktechnologie Pronunciation assessment & training by means of speech technology Helmer Strik – and many.

Slides:



Advertisements
Similar presentations
Excess Heat Production During D2 Diffusion Through Palladium
Advertisements

PowerPoint Tutorial 1 Creating a Presentation
Word Tutorial 2 Editing and Formatting a Document
Excel Tutorial 3 Working with Formulas and Functions
Objectives Create an action query to create a table
2400 Boston St Suite 407 Baltimore MD, (410) AP Optimization Building a World-Class Accounts Payable Operation By Brian Rosenberg.
Academic Literacy and Study Strategies for Secondary-level and College Readers: A Research Update International Reading Associations 2007 Annual Convention.
Scaling distributed search for diagnostics and prognostics applications Prof. Jim Austin Computer Science, University of York UK CEO Cybula Ltd.
Measurement of the Hard photodisintegration of a proton-pair in the 3 He nucleus. Running now in JLab, hall A E03-101:GamPP n p p Ishay Pomerantz, Tel-Aviv.
The Wiki and the Blog NIH Wiki Fair D. Calvin Andrus Chief Technology Officer Center for Mission Innovation Central Intelligence Agency 28 February 2007.
Rule 132 Medicaid Community Mental Health Service Program
Equilibria and Complexity: What now? Christos H. Papadimitriou UC Berkeley christos.
LAW February A 5/3 approximation algorithm for a hard case of stable marriage Rob Irving Computing Science Department University of Glasgow (joint.
3 october Brown easyBorrow (beta) Brown University Library October 2007.
Word Tutorial 1 Creating a Document
V3.0 12/04/ For Coral Systems With PRI Services Presented by Education Technology Services.
Corrective Feedback – pronunciation errors How effective it is in learning L2 oral communication Nguyễn Thị Tố Hạnh.
Elliott / October Understanding the Construct to be Assessed Stephen N. Elliott, PhD Learning Science Institute & Dept. of Special Education Vanderbilt.
Linked Lists: Locking, Lock- Free, and Beyond … Based on the companion slides for The Art of Multiprocessor Programming (Maurice Herlihy & Nir Shavit)
Italian white certificates scheme
Objectives Explore a structured range of data Freeze rows and columns
March 8, Dynamic Fault Tree analysis using Input/Output Interactive Markov Chains Hichem Boudali, Pepijn Crouzen, and Mariëlle Stoelinga. Formal.
Naam van de Auteur 7 januari 2008 The implementation results of the E- portfolio NL specification H-P Köhler 23 Oktober 2008.
Regional Agreements on Dermatological Extemporaneous Preparations Dermatologische Magistralrezeptur: Erfahrungen und Empfehlungen aus Holland FTTO RADA.
DISCO Development and Integration of Speech technology into Courseware for language learning Stevin project partners: CLST, UA, UTN, Polderland Radboud.
How to integrate automatic speech recognition (ASR) into CALL applications Helmer Strik Department of Linguistics Centre for Language and Speech Technology.
Sonar l TELE e-coaching application l de Overdracht bv © 2007 SONAR TELE e-coaching application.
Tertoolen/Bosch EECERA Many children, many voices Anja Tertoolen Wieke Bosch.
Madrid 7 -9 March FARO EU Kick-off Meeting Introduction to the project by Marta Pérez-Soba.
21 st Century Literacy Across Content Areas With Lin Kuzmich Senior Consultant Fall Symposium, Atlanta 2007.
IA/Spring Task Design EL441 Introduction to Teaching Young Learners Nilüfer Demirkan-Jones Session 3.
Word Tutorial 5 Working with Templates and Outlines
USSGL Board Meeting 4/26/20071 U.S. Treasury Zero Coupon Bonds (ZCB)
©SHRM 2007 HR’s EARLY WARNING SYSTEMS: Retention Strategies for the HR Professional Created by: SHRM’s Employee Relations Expertise Panel.
Broadband Adoption: Patterns, Behaviors, and Implications Presented to the New Jersey Connected Broadband Summit John B. Horrigan Associate Director for.
Helping Online Students Do Better Academically with Useful Learning & Study Strategies. Some recommended instructor activities to help students succeed.
SchedulingCS-502 Fall Scheduling (continued) The art and science of allocating the CPU and other resources to processes (Slides include materials.
Eunice Lumsden Eunice Lumsden The University of Northampton England A New Professional: Reflections on the Pilot Phase of the Early Years Professional.
Feza Baskaya – Anne Kakkonen - University of Tampere Second International Seminar on Subject Access to Information, Helsinki 30th November 2007.
Catia Cucchiarini Quantitative assessment of second language learners’ fluency in read and spontaneous speech Radboud University Nijmegen.
Sentence Durations and Accentedness Judgments ABSTRACT Talkers in a second language can frequently be identified as speaking with a foreign accent. It.
Why an objective intelligibility assessment ? Catherine Middag Jean-Pierre Martens Gwen Van Nuffelen Marc De Bodt.
CSD 5400 REHABILITATION PROCEDURES FOR THE HARD OF HEARING Auditory Training.
12.0 Computer-Assisted Language Learning (CALL) References: 1.“An Overview of Spoken Language Technology for Education”, Speech Communications, 51, pp ,
14: THE TEACHING OF GRAMMAR  Should grammar be taught?  When? How? Why?  Grammar teaching: Any strategies conducted in order to help learners understand,
Building a sentential model for automatic prosody evaluation Kyuchul Yoon School of English Language & Literature Yeungnam University Korea.
American Speechsounds How to Use the Program. AmericanSpeechsounds Why use American Speechsounds? Practice the problem sounds of American English Learn.
Using Technology to Teach Pronunciation A review of the research from Melike Yücel Eleonora Frigo Laurie Wayne Ling 578, Winter 2010, Dr. Arnold.
English Phonetics 许德华 许德华. Objectives of the Course This course is intended to help the students to improve their English pronunciation, including such.
New Acoustic-Phonetic Correlates Sorin Dusan and Larry Rabiner Center for Advanced Information Processing Rutgers University Piscataway,
Roman Kálecký UČO: Segmental features  Sounds  Speach Trainer 3D Suprasegmental features and accents  Speak English  SpeakAP  Accentuate!
Automatic Pronunciation Scoring of Specific Phone Segments for Language Instruction EuroSpeech 1997 Authors: Y. Kim, H. Franco, L. Neumeyer Presenter:
TEACHING PRONUNCIATION
Phone-Level Pronunciation Scoring and Assessment for Interactive Language Learning Speech Communication, 2000 Authors: S. M. Witt, S. J. Young Presenter:
Outline  I. Introduction  II. Reading fluency components  III. Experimental study  1) Method and participants  2) Testing materials  IV. Interpretation.
Prominent in English Teaching for Taiwan EFL Learning 指導教授 : 鍾榮富 高師大博士生 范春銀.
CHASING CHAllenging Speech training In Neurological patients by interactive Gaming Utrecht, November 12, 2015.
Dean Luo, Wentao Gu, Ruxin Luo and Lixin Wang
Sentence Durations and Accentedness Judgments
Teaching pronunciation
ASR-based corrective feedback on pronunciation: does it really work?
Dean Luo, Wentao Gu, Ruxin Luo and Lixin Wang
The sociocultural perspective to researching CAPT technology in the L2 speaking class for pronunciation training Moustafa Amrate Supervised by: Dr Irena.
SYSTEM APPROACH TO EDUCATION
Anastassia Loukina, Klaus Zechner, James Bruno, Beata Beigman Klebanov
Teaching prominence through kazoos
2017 APSIPA A Study on Landmark Detection Based on CTC and Its Application to Pronunciation Error Detection Chuanying Niu1, Jinsong Zhang1, Xuesong Yang2.
Presentation transcript:

Uitspraakevaluatie & training met behulp van spraaktechnologie Pronunciation assessment & training by means of speech technology Helmer Strik – and many others Centre for Language and Speech Technology (CLST) Radboud University Nijmegen, the Netherlands Radboud University Nijmegen

Leuven, Context ‘Deviant’ pronunciation (e.g., pathology, non-natives) & speech technology (applications) :  Assessment Diagnosis, monitoring  Training (therapy, learning) Speaking & listening; reading aloud  AAC (Augmentative & Alternative Communication) Improve communication

Radboud University Nijmegen Leuven, Our research  Past:  Fluency assessment - Temporal measures  CAPT: Computer Assisted Pronunciation Training  Pronunciation error detection  Recognition of dysarthric speech  Current, future:  OSTT: Ontwikkelcentrum voor Spraak- en Taaltechnologie ten behoeve van Spraak- en Taalpathologie en Revalidatietechnologie  Training & error detection, not only pronunciation, but also other (e.g. morpho-syntactic) aspects

Radboud University Nijmegen Leuven, Our research  Past:  Fluency assessment - Temporal measures  CAPT: Computer Assisted Pronunciation Training  Pronunciation error detection  Recognition of dysarthric speech  Current, future:  OSTT: Ontwikkelcentrum voor Spraak- en Taaltechnologie ten behoeve van Spraak- en Taalpathologie en Revalidatietechnologie  Training & error detection, not only pronunciation, but also other (e.g. morpho-syntactic) aspects

Radboud University Nijmegen Leuven, CAPT: Computer Assisted Pronunciation Training Pronunciation errors – detected automatically by means of Automatic Speech Recognition (ASR) → feedback  Question: ASR-based CAPT: Is it effective?  Goal: To study the effectiveness and possible advantages of ASR-based CAPT  Target users : Adult learners of Dutch with different L1's  Pedagogical goal : Improving segmental quality in pronunciation

Radboud University Nijmegen Leuven, Dutch CAPT: feedback Content: focus on problematic phonemes Criteria 1.Common across speakers of various L1’s 2.Perceptually salient 3.Frequent 4.Persistent 5.Robust for automatic detection (ASR) Result: 11 ‘targeted phonemes’: 9 vowels and 2 consonants

Radboud University Nijmegen Leuven, ‘targeted phonemes’ IPA symbolexample ////////toch, Scheveningen ////////hand, Helmer ////////pat /  :/ naam ////////pit ////////put ////////vuur ////////voer /  :/ deur /  / fijn /  / huis

Radboud University Nijmegen Leuven, Video (from Nieuwe Buren)

Radboud University Nijmegen Leuven,

Radboud University Nijmegen Leuven, Video: dialogue

Radboud University Nijmegen Leuven, Max. 3 times

Radboud University Nijmegen Leuven,

Radboud University Nijmegen Leuven,

Radboud University Nijmegen Leuven, Experiment: participants & training Regular teacher-fronted lessons: 4-6 hrs per week a)Experimental group (EXP): n=15 (10 F, 5 M) Dutch CAPT b)Control group 1 (NiBu): n=10 (4 F, 6 M) reduced version of Nieuwe Buren c)Control group 2 (noXT): n=5 (3 F, 2 M) no extra training Extra training: 4 weeks x 1 session 30’ – 60’ 1 class – 1 type of training

Radboud University Nijmegen Leuven, Experiment: testing 3 analyses: 1.Participants’ evaluations: questionnaires on system’s usability, accessibility, usefulness etc. 2.Global segmental quality: 6 experts rated stimuli on 10-point scale (pretest/posttest, phonetically balanced sentences) 3.In-depth analysis of segmental errors: expert annotations

Radboud University Nijmegen Leuven, Results: participants’ evaluations Positive reactions Enjoyed working with the system Believed in the usefulness of the system

Radboud University Nijmegen Leuven, Results: Global segmental quality All 3 groups improve (mean improvement) EXP improved most

Radboud University Nijmegen Leuven, In-depth analysis of segmental errors

Radboud University Nijmegen Leuven, Conclusions  Goal: To study the effectiveness and possible advantages of ASR-based CAPT  Question: ASR-based CAPT: Is it effective? Answer: Yes! It is effective in improving the pronunciation of targeted phonemes.  Advantages : ASR-based CAPT can provide automatic, instantaneous, individual feedback on pronunciation in a private environment.

Radboud University Nijmegen Leuven, Video: pronouncing words

Radboud University Nijmegen Leuven, Error detection Detection of pronunciation errors  Goodness Of Pronunciation (GOP) oSilke Witt & Steve Young  Acoustic-phonetic features (APF) oKhiet Truong et al.  Goal: improve error detection

Radboud University Nijmegen Leuven,

Radboud University Nijmegen Leuven, Goodness Of Pronunciation (GOP): Accuracy 15 participants 2174 target phones AcceptRejectTotal CorrectCA: 59.5%CR: 26.5%C: 86.0% FalseFA: 9.2%FR: 4.8%F: 14.0%

Radboud University Nijmegen Leuven, Acoustic-phonetic features (APF) Selection of segmental pronunciation errors: /A/ mispronounced as /a:/ (man - maan) /Y/ mispronounced as /u/ or /y/ (tut – toet or tuut) /x/ mispronounced as /k/ or /g/ (gat – kat or /g/at)

Radboud University Nijmegen Leuven, Amplitude Rate Of Rise (ROR)

Radboud University Nijmegen Leuven, Height of the highest ROR peak (‘ROR’) 1 amplitude measurement before the ROR peak (‘i1’) 3 amplitude measurements after the ROR peak (‘i2’, ‘i3’, ‘i4’) Duration (‘rawdur’ or ‘normdur’ or not used at all ‘nodur’)

Radboud University Nijmegen Leuven,

Radboud University Nijmegen Leuven, Error detection Goodness Of Pronunciation (GOP):  One general method for all sounds  Error specific knowledge is not used Acoustic-phonetic features (APF)  Error specific knowledge is used  Works well  How to generalize? (artic. + other features) Combination? Other approaches, e.g. post. prob’s (ANN)?

Radboud University Nijmegen Leuven,

Radboud University Nijmegen Leuven,

Radboud University Nijmegen Leuven, Dutch CAPT Gender-specific, Dutch & English version. 4 units, each containing: 1 video (from Nieuwe Buren) with real-life + amusing situations + ca. 30 exercises based on video: dialogues, question- answer, minimal pairs, word repetition Sequential, constrained navigation: min. one attempt needed to proceed to next exercise, maximum 3

Radboud University Nijmegen Leuven, Results: reliability global ratings Cronbach’s α: Intrarater: 0.94 – 1.00 Interrater:

Radboud University Nijmegen Leuven,

Radboud University Nijmegen Leuven, Results: Global ratings

Radboud University Nijmegen Leuven, Possible improvements  Increase sample size (more participants)  Increase training intensity (more training)  Match training groups: L1’s, proficiency, etc.  Give feedback on more phonemes More targeted systems for fixed L1-L2 pairs.  Give feedback on suprasegmentals  Improve error detection?

Radboud University Nijmegen Leuven, Error detection Pronunciation errors  11 ‘problematic sounds’: 9 V + 2 C  Goal: give feedback on more sounds Morpho-syntactic errors  maak / maakt / maken oIk maak oHij/zij maakt oWij maken  Goal: also give feedback on morpho-syntactic aspects

Radboud University Nijmegen Leuven, Goodness Of Pronunciation (GOP) GOP has been applied in the exp. system. The exp. system was effective. Evaluate GOP  Correct vs. errors  Patterns  Pros & cons  Improve

Radboud University Nijmegen Leuven,

Radboud University Nijmegen Leuven,

Radboud University Nijmegen Leuven,

Radboud University Nijmegen Leuven, Training = DL2N1-Nat Test = DL2N1-Nat Training = DL2N1-NN Test = DL2N1-NN Results method II (LDA) /x/ vs /k/

Radboud University Nijmegen Leuven, ‘targeted phonemes’ /  /, /  /, /  /, /  /, /  /, /  :/, /  /, /  /, /  /, /  :/, /  /