Automatic dialect detection in Arabic broadcast speech

Ahmed Ali, Najim Dehak, Patrick Cardinal, Sameer Khurana, Sree Harsha Yella, James Glass, Peter Bell, Steve Renals

Research output: Contribution to journalArticle

24 Citations (Scopus)

Abstract

In this paper, we investigate different approaches for dialect identification in Arabic broadcast speech. These methods are based on phonetic and lexical features obtained from a speech recognition system, and bottleneck features using the i-vector framework. We studied both generative and discriminative classifiers, and we combined these features using a multi-class Support Vector Machine (SVM). We validated our results on an Arabic/English language identification task, with an accuracy of 100%. We also evaluated these features in a binary classifier to discriminate between Modern Standard Arabic (MSA) and Dialectal Arabic, with an accuracy of 100%. We further reported results using the proposed methods to discriminate between the five most widely used dialects of Arabic: namely Egyptian, Gulf, Levantine, North African, and MSA, with an accuracy of 59.2%. We discuss dialect identification errors in the context of dialect code-switching between Dialectal Arabic and MSA, and compare the error pattern between manually labeled data, and the output from our classifier. All the data used on our experiments have been released to the public as a language identification corpus.

Original languageEnglish
Pages (from-to)2934-2938
Number of pages5
JournalUnknown Journal
Volume08-12-September-2016
DOIs
Publication statusPublished - 2016

Fingerprint

Language Identification
Broadcast
Classifiers
Classifier
Speech analysis
Multi-class
Speech Recognition
Speech recognition
Support vector machines
Support Vector Machine
Binary
Output
Experiment
Standards
Speech
automatic detection
speech
experiment
Experiments
method

Keywords

  • Dialect Identification
  • Vector Space Modelling

ASJC Scopus subject areas

  • Language and Linguistics
  • Human-Computer Interaction
  • Signal Processing
  • Software
  • Modelling and Simulation

Cite this

Ali, A., Dehak, N., Cardinal, P., Khurana, S., Yella, S. H., Glass, J., ... Renals, S. (2016). Automatic dialect detection in Arabic broadcast speech. Unknown Journal, 08-12-September-2016, 2934-2938. https://doi.org/10.21437/Interspeech.2016-1297

Automatic dialect detection in Arabic broadcast speech. / Ali, Ahmed; Dehak, Najim; Cardinal, Patrick; Khurana, Sameer; Yella, Sree Harsha; Glass, James; Bell, Peter; Renals, Steve.

In: Unknown Journal, Vol. 08-12-September-2016, 2016, p. 2934-2938.

Research output: Contribution to journalArticle

Ali, A, Dehak, N, Cardinal, P, Khurana, S, Yella, SH, Glass, J, Bell, P & Renals, S 2016, 'Automatic dialect detection in Arabic broadcast speech', Unknown Journal, vol. 08-12-September-2016, pp. 2934-2938. https://doi.org/10.21437/Interspeech.2016-1297
Ali A, Dehak N, Cardinal P, Khurana S, Yella SH, Glass J et al. Automatic dialect detection in Arabic broadcast speech. Unknown Journal. 2016;08-12-September-2016:2934-2938. https://doi.org/10.21437/Interspeech.2016-1297
Ali, Ahmed ; Dehak, Najim ; Cardinal, Patrick ; Khurana, Sameer ; Yella, Sree Harsha ; Glass, James ; Bell, Peter ; Renals, Steve. / Automatic dialect detection in Arabic broadcast speech. In: Unknown Journal. 2016 ; Vol. 08-12-September-2016. pp. 2934-2938.
@article{4077910337d54b3495013f57724ac1ea,
title = "Automatic dialect detection in Arabic broadcast speech",
abstract = "In this paper, we investigate different approaches for dialect identification in Arabic broadcast speech. These methods are based on phonetic and lexical features obtained from a speech recognition system, and bottleneck features using the i-vector framework. We studied both generative and discriminative classifiers, and we combined these features using a multi-class Support Vector Machine (SVM). We validated our results on an Arabic/English language identification task, with an accuracy of 100{\%}. We also evaluated these features in a binary classifier to discriminate between Modern Standard Arabic (MSA) and Dialectal Arabic, with an accuracy of 100{\%}. We further reported results using the proposed methods to discriminate between the five most widely used dialects of Arabic: namely Egyptian, Gulf, Levantine, North African, and MSA, with an accuracy of 59.2{\%}. We discuss dialect identification errors in the context of dialect code-switching between Dialectal Arabic and MSA, and compare the error pattern between manually labeled data, and the output from our classifier. All the data used on our experiments have been released to the public as a language identification corpus.",
keywords = "Dialect Identification, Vector Space Modelling",
author = "Ahmed Ali and Najim Dehak and Patrick Cardinal and Sameer Khurana and Yella, {Sree Harsha} and James Glass and Peter Bell and Steve Renals",
year = "2016",
doi = "10.21437/Interspeech.2016-1297",
language = "English",
volume = "08-12-September-2016",
pages = "2934--2938",
journal = "JAPCA",
issn = "1073-161X",
publisher = "Taylor and Francis Ltd.",

}

TY - JOUR

T1 - Automatic dialect detection in Arabic broadcast speech

AU - Ali, Ahmed

AU - Dehak, Najim

AU - Cardinal, Patrick

AU - Khurana, Sameer

AU - Yella, Sree Harsha

AU - Glass, James

AU - Bell, Peter

AU - Renals, Steve

PY - 2016

Y1 - 2016

N2 - In this paper, we investigate different approaches for dialect identification in Arabic broadcast speech. These methods are based on phonetic and lexical features obtained from a speech recognition system, and bottleneck features using the i-vector framework. We studied both generative and discriminative classifiers, and we combined these features using a multi-class Support Vector Machine (SVM). We validated our results on an Arabic/English language identification task, with an accuracy of 100%. We also evaluated these features in a binary classifier to discriminate between Modern Standard Arabic (MSA) and Dialectal Arabic, with an accuracy of 100%. We further reported results using the proposed methods to discriminate between the five most widely used dialects of Arabic: namely Egyptian, Gulf, Levantine, North African, and MSA, with an accuracy of 59.2%. We discuss dialect identification errors in the context of dialect code-switching between Dialectal Arabic and MSA, and compare the error pattern between manually labeled data, and the output from our classifier. All the data used on our experiments have been released to the public as a language identification corpus.

AB - In this paper, we investigate different approaches for dialect identification in Arabic broadcast speech. These methods are based on phonetic and lexical features obtained from a speech recognition system, and bottleneck features using the i-vector framework. We studied both generative and discriminative classifiers, and we combined these features using a multi-class Support Vector Machine (SVM). We validated our results on an Arabic/English language identification task, with an accuracy of 100%. We also evaluated these features in a binary classifier to discriminate between Modern Standard Arabic (MSA) and Dialectal Arabic, with an accuracy of 100%. We further reported results using the proposed methods to discriminate between the five most widely used dialects of Arabic: namely Egyptian, Gulf, Levantine, North African, and MSA, with an accuracy of 59.2%. We discuss dialect identification errors in the context of dialect code-switching between Dialectal Arabic and MSA, and compare the error pattern between manually labeled data, and the output from our classifier. All the data used on our experiments have been released to the public as a language identification corpus.

KW - Dialect Identification

KW - Vector Space Modelling

UR - http://www.scopus.com/inward/record.url?scp=84994246070&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84994246070&partnerID=8YFLogxK

U2 - 10.21437/Interspeech.2016-1297

DO - 10.21437/Interspeech.2016-1297

M3 - Article

AN - SCOPUS:84994246070

VL - 08-12-September-2016

SP - 2934

EP - 2938

JO - JAPCA

JF - JAPCA

SN - 1073-161X

ER -