Comparing Different Methods for Disfluency Structure Detection

Authors Henrique Medeiros, Fernando Batista, Helena Moniz, Isabel Trancoso, Luis Nunes



PDF
Thumbnail PDF

File

OASIcs.SLATE.2013.259.pdf
  • Filesize: 408 kB
  • 11 pages

Document Identifiers

Author Details

Henrique Medeiros
Fernando Batista
Helena Moniz
Isabel Trancoso
Luis Nunes

Cite As Get BibTex

Henrique Medeiros, Fernando Batista, Helena Moniz, Isabel Trancoso, and Luis Nunes. Comparing Different Methods for Disfluency Structure Detection. In 2nd Symposium on Languages, Applications and Technologies. Open Access Series in Informatics (OASIcs), Volume 29, pp. 259-269, Schloss Dagstuhl – Leibniz-Zentrum für Informatik (2013) https://doi.org/10.4230/OASIcs.SLATE.2013.259

Abstract

This paper presents a number of experiments focusing on assessing
the performance of different machine learning methods on the identification of disfluencies and their distinct structural regions over speech data. Several machine learning methods have been applied, namely Naive Bayes, Logistic Regression, Classification and Regression Trees (CARTs), J48 and Multilayer Perceptron. Our experiments show that CARTs outperform the other methods on the identification of the distinct structural disfluent regions. Reported experiments are based on audio segmentation and prosodic features, calculated from a corpus of university lectures in European Portuguese, containing about 32h of speech and about 7.7% of disfluencies. The set of features automatically extracted from the forced alignment corpus proved to be discriminant of the regions contained in the production of a disfluency. This work shows that
using fully automatic prosodic features, disfluency structural regions
can be reliably identified using CARTs, where the best results achieved correspond to 81.5% precision, 27.6% recall, and 41.2% F-measure. The best results concern the detection of the interregnum, followed by the detection of the interruption point.

Subject Classification

Keywords
  • Machine learning
  • speech processing
  • prosodic features
  • automatic detection of disfluencies

Metrics

  • Access Statistics
  • Total Accesses (updated on a weekly basis)
    0
    PDF Downloads
Questions / Remarks / Feedback
X

Feedback for Dagstuhl Publishing


Thanks for your feedback!

Feedback submitted

Could not send message

Please try again later or send an E-mail