Automatic prosodic tone choice classification with Brazil’s intonation model

David O. Johnson, Okim Kang

Research output: Contribution to journalArticle

7 Citations (Scopus)

Abstract

This paper examines the performance of automatically classifying five tone choices (i.e., falling, rising, rising-falling, falling-rising, and neutral) of Brazil’s intonation model. We tested two machine learning classifiers (neural network and boosting ensemble) in two configurations (multi-class and pairwise coupling) and a rule-based classifier. Three sets of acoustic features built from the TILT and Bézier pitch contour models and a new four-point pitch contour model we introduced here were investigated. Tone choices are one of the key elements of Brazil’s prosodic intonation model. We found the rule-based classifier, which was built on our four-point model, achieved better results than the others with an accuracy of 75.1 % and a Cohen’s kappa coefficient of 0.73. This research proves that it is possible to classify tone choices with an accuracy reaching close to the percentage of agreement between two human analysts. The findings further concluded that our four-point model was better for classifying Brazil’s tone choices than both of the TILT or Bézier models.

Original languageEnglish (US)
Pages (from-to)95-109
Number of pages15
JournalInternational Journal of Speech Technology
Volume19
Issue number1
DOIs
StatePublished - Mar 1 2016

Fingerprint

Classifiers
Intonation
neural network
acoustics
Learning systems
Acoustics
Neural networks
learning
performance
Classifier
Pitch Contour
Neural Networks
Ensemble
Machine Learning

Keywords

  • Brazil’s prosodic intonation model
  • Bézier model
  • Machine learning
  • TILT model
  • ToBI
  • Tone choice classification

ASJC Scopus subject areas

  • Computer Vision and Pattern Recognition
  • Software
  • Human-Computer Interaction
  • Language and Linguistics
  • Linguistics and Language

Cite this

Automatic prosodic tone choice classification with Brazil’s intonation model. / Johnson, David O.; Kang, Okim.

In: International Journal of Speech Technology, Vol. 19, No. 1, 01.03.2016, p. 95-109.

Research output: Contribution to journalArticle

@article{f5d71b33d37849bc9c117272d0b48596,
title = "Automatic prosodic tone choice classification with Brazil’s intonation model",
abstract = "This paper examines the performance of automatically classifying five tone choices (i.e., falling, rising, rising-falling, falling-rising, and neutral) of Brazil’s intonation model. We tested two machine learning classifiers (neural network and boosting ensemble) in two configurations (multi-class and pairwise coupling) and a rule-based classifier. Three sets of acoustic features built from the TILT and B{\'e}zier pitch contour models and a new four-point pitch contour model we introduced here were investigated. Tone choices are one of the key elements of Brazil’s prosodic intonation model. We found the rule-based classifier, which was built on our four-point model, achieved better results than the others with an accuracy of 75.1 {\%} and a Cohen’s kappa coefficient of 0.73. This research proves that it is possible to classify tone choices with an accuracy reaching close to the percentage of agreement between two human analysts. The findings further concluded that our four-point model was better for classifying Brazil’s tone choices than both of the TILT or B{\'e}zier models.",
keywords = "Brazil’s prosodic intonation model, B{\'e}zier model, Machine learning, TILT model, ToBI, Tone choice classification",
author = "Johnson, {David O.} and Okim Kang",
year = "2016",
month = "3",
day = "1",
doi = "10.1007/s10772-015-9327-z",
language = "English (US)",
volume = "19",
pages = "95--109",
journal = "International Journal of Speech Technology",
issn = "1381-2416",
publisher = "Springer Netherlands",
number = "1",

}

TY - JOUR

T1 - Automatic prosodic tone choice classification with Brazil’s intonation model

AU - Johnson, David O.

AU - Kang, Okim

PY - 2016/3/1

Y1 - 2016/3/1

N2 - This paper examines the performance of automatically classifying five tone choices (i.e., falling, rising, rising-falling, falling-rising, and neutral) of Brazil’s intonation model. We tested two machine learning classifiers (neural network and boosting ensemble) in two configurations (multi-class and pairwise coupling) and a rule-based classifier. Three sets of acoustic features built from the TILT and Bézier pitch contour models and a new four-point pitch contour model we introduced here were investigated. Tone choices are one of the key elements of Brazil’s prosodic intonation model. We found the rule-based classifier, which was built on our four-point model, achieved better results than the others with an accuracy of 75.1 % and a Cohen’s kappa coefficient of 0.73. This research proves that it is possible to classify tone choices with an accuracy reaching close to the percentage of agreement between two human analysts. The findings further concluded that our four-point model was better for classifying Brazil’s tone choices than both of the TILT or Bézier models.

AB - This paper examines the performance of automatically classifying five tone choices (i.e., falling, rising, rising-falling, falling-rising, and neutral) of Brazil’s intonation model. We tested two machine learning classifiers (neural network and boosting ensemble) in two configurations (multi-class and pairwise coupling) and a rule-based classifier. Three sets of acoustic features built from the TILT and Bézier pitch contour models and a new four-point pitch contour model we introduced here were investigated. Tone choices are one of the key elements of Brazil’s prosodic intonation model. We found the rule-based classifier, which was built on our four-point model, achieved better results than the others with an accuracy of 75.1 % and a Cohen’s kappa coefficient of 0.73. This research proves that it is possible to classify tone choices with an accuracy reaching close to the percentage of agreement between two human analysts. The findings further concluded that our four-point model was better for classifying Brazil’s tone choices than both of the TILT or Bézier models.

KW - Brazil’s prosodic intonation model

KW - Bézier model

KW - Machine learning

KW - TILT model

KW - ToBI

KW - Tone choice classification

UR - http://www.scopus.com/inward/record.url?scp=84957432026&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84957432026&partnerID=8YFLogxK

U2 - 10.1007/s10772-015-9327-z

DO - 10.1007/s10772-015-9327-z

M3 - Article

VL - 19

SP - 95

EP - 109

JO - International Journal of Speech Technology

JF - International Journal of Speech Technology

SN - 1381-2416

IS - 1

ER -