Publication Details

Automatic Generation Of Pronunciation Dictionaries Based On Diarization

JANDA Miloš. Automatic Generation Of Pronunciation Dictionaries Based On Diarization. In: Proceedings of the 19th Conference Student EEICT 2013. Brno: Brno University of Technology, 2013, pp. 228-232. ISBN 978-80-214-4695-3. Available from: http://www.feec.vutbr.cz/EEICT/2013/sbornik/03doktorskeprojekty/09grafikaamultimedia/01-xjanda16.pdf
Czech title
Automatické generování výslovnostních slovníků založené na diarizaci
Type
conference paper
Language
english
Authors
Janda Miloš, Ing. (DCGM FIT BUT)
URL
Keywords

speech recognition, LVCSR, vocabulary, pronunciation dictionary, lexicon, diarization

Abstract

This article is about an automatic generation of pronunciation dictionaries, which are based on diarization approach.

Annotation

This paper presents the results of eight speech recognizers with automatically generated pronunciation dictionaries. Diarization approach, typically used in speaker recognition, could be modified for purpose of automatic generation of dictionaries, targeting the low resource languages where acquisition of hand-crafted pronunciation dictionary is time- and cost-consuming or impossi- ble. Experiments on GlobalPhone database show that diarization approach is suitable alternative for grapheme-based method.

Published
2013
Pages
228-232
Proceedings
Proceedings of the 19th Conference Student EEICT 2013
Conference
Student EEICT 2013, Brno, CZ
ISBN
978-80-214-4695-3
Publisher
Brno University of Technology
Place
Brno, CZ
BibTeX 
@INPROCEEDINGS{FITPUB10342,
   author = "Milo\v{s} Janda",
   title = "Automatic Generation Of Pronunciation Dictionaries Based On Diarization",
   pages = "228--232",
   booktitle = "Proceedings of the 19th Conference Student EEICT 2013",
   year = 2013,
   location = "Brno, CZ",
   publisher = "Brno University of Technology",
   ISBN = "978-80-214-4695-3",
   language = "english",
   url = "https://www.fit.vut.cz/research/publication/10342"
}
Back to top