Publication Details

Grapheme Based Speech Recognition

JANDA Miloš. Grapheme Based Speech Recognition. In: Proceedings of the 18th Conference STUDENT EEICT 2012. Brno: Brno University of Technology, 2012, pp. 441-445. ISBN 978-80-214-4460-7. Available from: http://www.feec.vutbr.cz/EEICT/2012/sbornik/03doktorskeprojekty/09grafikaamultimedia/02-xjanda16.pdf
Czech title
Rozpoznávání řeči založené na grafémech
Type
conference paper
Language
english
Authors
Janda Miloš, Ing. (DCGM FIT BUT)
URL
Keywords

speech recognition, LVCSR, ASR, grapheme, phoneme, low-resource languages

Abstract

This article describes the results of grapheme-based speech recognition for eight languages. We used this approach due to the situation of low resource languages.

Annotation

This article presents the results of grapheme-based speech recognition for eight languages. The need for this approach arises in situation of low resource languages, where obtaining a pronunciation dictionary is time- and cost-consuming or impossible. In such scenarios, usage of grapheme dictionaries is the most simplest and straight-forward. The paper describes the process of automatic generation of pronunciation dictionaries with emphasis on the expansion of numbers. Experiments on GlobalPhone database show that grapheme-based systems have results comparable to the phonemebased ones, especially for phonetic languages.

Published
2012
Pages
441-445
Proceedings
Proceedings of the 18th Conference STUDENT EEICT 2012
Conference
Student EEICT 2012, Brno, CZ
ISBN
978-80-214-4460-7
Publisher
Brno University of Technology
Place
Brno, CZ
BibTeX 
@INPROCEEDINGS{FITPUB10043,
   author = "Milo\v{s} Janda",
   title = "Grapheme Based Speech Recognition",
   pages = "441--445",
   booktitle = "Proceedings of the 18th Conference STUDENT EEICT 2012",
   year = 2012,
   location = "Brno, CZ",
   publisher = "Brno University of Technology",
   ISBN = "978-80-214-4460-7",
   language = "english",
   url = "https://www.fit.vut.cz/research/publication/10043"
}
Back to top