2026-07-26T02:38:15Z https://drops.dagstuhl.de/oai

oai:drops-oai.dagstuhl.de:4042 2024-03-06T10:29:13Z ddc:004 open_access

Comparing Different Methods for Disfluency Structure Detection Medeiros, Henrique Batista, Fernando Moniz, Helena Trancoso, Isabel Nunes, Luis Machine learning speech processing prosodic features automatic detection of disfluencies 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. Schloss Dagstuhl – Leibniz-Zentrum für Informatik Henrique Medeiros and Fernando Batista and Helena Moniz and Isabel Trancoso and Luis Nunes 2013 Is Part Of OASIcs, Volume 29, 2nd Symposium on Languages, Applications and Technologies (2013) InProceedings Text doc-type:ResearchArticle publishedVersion application/pdf doi:10.4230/OASIcs.SLATE.2013.259 urn:nbn:de:0030-drops-40420 https://drops.dagstuhl.de/entities/document/10.4230/OASIcs.SLATE.2013.259 eng https://creativecommons.org/licenses/by/3.0/legalcode