Morpheme-based Language Models for Improving the Speech Recognition of Arabic Dialects

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

Standard

Morpheme-based Language Models for Improving the Speech Recognition of Arabic Dialects. / Almeman, Khalid; Lee, Mark.

Proceedings of the 5th International Conference on Arabic Language Processing: (CITALA '14) . CITALA, 2014. p. 49-56 19.

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

Harvard

Almeman, K & Lee, M 2014, Morpheme-based Language Models for Improving the Speech Recognition of Arabic Dialects. in Proceedings of the 5th International Conference on Arabic Language Processing: (CITALA '14) ., 19, CITALA, pp. 49-56, 5th International Conference on Arabic Language Processing (CITALA '14) , Oujda, Morocco, 26/11/14.

APA

Almeman, K., & Lee, M. (2014). Morpheme-based Language Models for Improving the Speech Recognition of Arabic Dialects. In Proceedings of the 5th International Conference on Arabic Language Processing: (CITALA '14) (pp. 49-56). [19] CITALA.

Vancouver

Almeman K, Lee M. Morpheme-based Language Models for Improving the Speech Recognition of Arabic Dialects. In Proceedings of the 5th International Conference on Arabic Language Processing: (CITALA '14) . CITALA. 2014. p. 49-56. 19

Author

Almeman, Khalid ; Lee, Mark. / Morpheme-based Language Models for Improving the Speech Recognition of Arabic Dialects. Proceedings of the 5th International Conference on Arabic Language Processing: (CITALA '14) . CITALA, 2014. pp. 49-56

Bibtex

@inproceedings{7fbefcd742d34a369425f3ef471c6529,
title = "Morpheme-based Language Models for Improving the Speech Recognition of Arabic Dialects",
abstract = "In this paper, innovative experiments were done toimprove the Language Models (LMs) for three parallel dialects. Ineach dialect, two different LMs were produced: a closed domainLM and an open domain LM. The methodology of the secondpart of the multi dialect morphology analyser, involved retrievalof web frequencies for different parts of a word; this methodologywas modified and then used to extract the three suggested formsof the word; stem alone, prefix+stem and stem+suffix. Six resultswere then extracted per dialect, giving a total of eighteen results.All the experiments yielded positive results, between 0.5% to6.8% in WERs.",
author = "Khalid Almeman and Mark Lee",
year = "2014",
month = nov
language = "English",
pages = "49--56",
booktitle = "Proceedings of the 5th International Conference on Arabic Language Processing",
publisher = "CITALA",
note = "5th International Conference on Arabic Language Processing (CITALA '14) ; Conference date: 26-11-2014 Through 27-11-2014",

}

RIS

TY - GEN

T1 - Morpheme-based Language Models for Improving the Speech Recognition of Arabic Dialects

AU - Almeman, Khalid

AU - Lee, Mark

PY - 2014/11

Y1 - 2014/11

N2 - In this paper, innovative experiments were done toimprove the Language Models (LMs) for three parallel dialects. Ineach dialect, two different LMs were produced: a closed domainLM and an open domain LM. The methodology of the secondpart of the multi dialect morphology analyser, involved retrievalof web frequencies for different parts of a word; this methodologywas modified and then used to extract the three suggested formsof the word; stem alone, prefix+stem and stem+suffix. Six resultswere then extracted per dialect, giving a total of eighteen results.All the experiments yielded positive results, between 0.5% to6.8% in WERs.

AB - In this paper, innovative experiments were done toimprove the Language Models (LMs) for three parallel dialects. Ineach dialect, two different LMs were produced: a closed domainLM and an open domain LM. The methodology of the secondpart of the multi dialect morphology analyser, involved retrievalof web frequencies for different parts of a word; this methodologywas modified and then used to extract the three suggested formsof the word; stem alone, prefix+stem and stem+suffix. Six resultswere then extracted per dialect, giving a total of eighteen results.All the experiments yielded positive results, between 0.5% to6.8% in WERs.

UR - http://www.citala.org/citala2014/acceptedpapers.html

UR - http://www.citala.org/icalp2017/#preEd

M3 - Conference contribution

SP - 49

EP - 56

BT - Proceedings of the 5th International Conference on Arabic Language Processing

PB - CITALA

T2 - 5th International Conference on Arabic Language Processing (CITALA '14)

Y2 - 26 November 2014 through 27 November 2014

ER -