Unsupervised prosodic break detection in Mandarin speech

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

We propose that, in Mandarin speech, an automatic prosodic break detector can be trained without any prosodically labeled training data. We use only lexical and acoustic cues to create a small labeled training set, then use semi-supervised learning to train a prosodic break detector. A generative mixture model is proposed as the learning algorithm that can learn with both labeled and unlabeled data. The experiments in both English and Mandarin corpus verify our algorithm.

Original languageEnglish (US)
Title of host publicationProceedings of the 4th International Conference on Speech Prosody
PublisherInternational Speech Communications Association
Pages165-168
Number of pages4
ISBN (Print)9780616220030
StatePublished - Jan 1 2008
Event4th International Conference on Speech Prosody 2008, SP 2008 - Campinas, Brazil
Duration: May 6 2008May 9 2008

Publication series

NameProceedings of the 4th International Conference on Speech Prosody, SP 2008

Other

Other4th International Conference on Speech Prosody 2008, SP 2008
CountryBrazil
CityCampinas
Period5/6/085/9/08

ASJC Scopus subject areas

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

Fingerprint Dive into the research topics of 'Unsupervised prosodic break detection in Mandarin speech'. Together they form a unique fingerprint.

  • Cite this

    Huang, J. T., Hasegawa-Johnson, M. A., & Shih, C. (2008). Unsupervised prosodic break detection in Mandarin speech. In Proceedings of the 4th International Conference on Speech Prosody (pp. 165-168). (Proceedings of the 4th International Conference on Speech Prosody, SP 2008). International Speech Communications Association.