Edit disfluency detection and correction using a cleanup language model and an alignment model

Jui Feng Yeh, Chung-Hsien Wu

Research output: Contribution to journalArticle

14 Citations (Scopus)

Abstract

This investigation presents a novel approach to detecting and correcting the edit disfluency in spontaneous speech. Hypothesis testing using acoustic features is first adopted to detect potential interruption points (IPs) in the input speech. The word order of the cleanup utterance is then cleaned up based on the potential IPs using a class-based cleanup language model, the dclctablc region and the correction are aligned using an alignment model. Finally, log linear weighting is applied to optimize the performance. Using the acoustic features, the IP detection rate is significantly improved especially in recall rate. Based on the positions of the potential IPs, the cleanup language model and the alignment model are able to detect and correct the edit disflucncy efficiently. Experimental results demonstrate that the proposed approach has achieved error rates of 0.33 and 0.21 for IP detection and edit word deletion, respectively.

Original languageEnglish
Pages (from-to)1574-1583
Number of pages10
JournalIEEE Transactions on Audio, Speech and Language Processing
Volume14
Issue number5
DOIs
Publication statusPublished - 2006 Sep 1

Fingerprint

interruption
alignment
Acoustics
deletion
acoustics
Testing

All Science Journal Classification (ASJC) codes

  • Acoustics and Ultrasonics
  • Electrical and Electronic Engineering

Cite this

@article{ffa72c4c7c3e4369bd0f1b9f3e1b04f7,
title = "Edit disfluency detection and correction using a cleanup language model and an alignment model",
abstract = "This investigation presents a novel approach to detecting and correcting the edit disfluency in spontaneous speech. Hypothesis testing using acoustic features is first adopted to detect potential interruption points (IPs) in the input speech. The word order of the cleanup utterance is then cleaned up based on the potential IPs using a class-based cleanup language model, the dclctablc region and the correction are aligned using an alignment model. Finally, log linear weighting is applied to optimize the performance. Using the acoustic features, the IP detection rate is significantly improved especially in recall rate. Based on the positions of the potential IPs, the cleanup language model and the alignment model are able to detect and correct the edit disflucncy efficiently. Experimental results demonstrate that the proposed approach has achieved error rates of 0.33 and 0.21 for IP detection and edit word deletion, respectively.",
author = "Yeh, {Jui Feng} and Chung-Hsien Wu",
year = "2006",
month = "9",
day = "1",
doi = "10.1109/TASL.2006.878267",
language = "English",
volume = "14",
pages = "1574--1583",
journal = "IEEE Transactions on Speech and Audio Processing",
issn = "1558-7916",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "5",

}

Edit disfluency detection and correction using a cleanup language model and an alignment model. / Yeh, Jui Feng; Wu, Chung-Hsien.

In: IEEE Transactions on Audio, Speech and Language Processing, Vol. 14, No. 5, 01.09.2006, p. 1574-1583.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Edit disfluency detection and correction using a cleanup language model and an alignment model

AU - Yeh, Jui Feng

AU - Wu, Chung-Hsien

PY - 2006/9/1

Y1 - 2006/9/1

N2 - This investigation presents a novel approach to detecting and correcting the edit disfluency in spontaneous speech. Hypothesis testing using acoustic features is first adopted to detect potential interruption points (IPs) in the input speech. The word order of the cleanup utterance is then cleaned up based on the potential IPs using a class-based cleanup language model, the dclctablc region and the correction are aligned using an alignment model. Finally, log linear weighting is applied to optimize the performance. Using the acoustic features, the IP detection rate is significantly improved especially in recall rate. Based on the positions of the potential IPs, the cleanup language model and the alignment model are able to detect and correct the edit disflucncy efficiently. Experimental results demonstrate that the proposed approach has achieved error rates of 0.33 and 0.21 for IP detection and edit word deletion, respectively.

AB - This investigation presents a novel approach to detecting and correcting the edit disfluency in spontaneous speech. Hypothesis testing using acoustic features is first adopted to detect potential interruption points (IPs) in the input speech. The word order of the cleanup utterance is then cleaned up based on the potential IPs using a class-based cleanup language model, the dclctablc region and the correction are aligned using an alignment model. Finally, log linear weighting is applied to optimize the performance. Using the acoustic features, the IP detection rate is significantly improved especially in recall rate. Based on the positions of the potential IPs, the cleanup language model and the alignment model are able to detect and correct the edit disflucncy efficiently. Experimental results demonstrate that the proposed approach has achieved error rates of 0.33 and 0.21 for IP detection and edit word deletion, respectively.

UR - http://www.scopus.com/inward/record.url?scp=34047266604&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=34047266604&partnerID=8YFLogxK

U2 - 10.1109/TASL.2006.878267

DO - 10.1109/TASL.2006.878267

M3 - Article

AN - SCOPUS:34047266604

VL - 14

SP - 1574

EP - 1583

JO - IEEE Transactions on Speech and Audio Processing

JF - IEEE Transactions on Speech and Audio Processing

SN - 1558-7916

IS - 5

ER -