Přejít na obsah

Detail publikace

Citace

Ircing, P. and Psutka, J. : Fitting class-based language models into weighted finite-state transducer framework . EUROSPEECH 2003 PROCEEDINGS, p. 1873-1876, ISCA, Geneva, 2003.

Abstrakt

In our paper we propose a general way of incorporating class-based language models with many-to-many word-to-class mapping into the finite-state transducer (FST) framework. Since class-based models alone usually do not improve the recognition accuracy, we also present a method for an efficient language model combination. An example of a word-to-class mapping based on morphological tags is also given. Several word-based and tag-based language models are tested in the task of transcribing Czech broadcast news. Results show that class-based models help to achieve a moderate improvement in recognition accuracy.

Detail publikace

Název: Fitting class-based language models into weighted finite-state transducer framework
Autor: Ircing, P. ; Psutka, J.
Jazyk publikace: anglicky
Datum vydání: 1.9.2003
Rok vydání: 2003
Typ publikace: Stať ve sborníku
Název časopisu / knihy: EUROSPEECH 2003 PROCEEDINGS
Strana: 1873 - 1876
Nakladatel: ISCA
Místo vydání: Geneva
Datum: 1.9.2003 - 4.9.2003
/ /

Klíčová slova

language modeling, finite-state automata, speech recognition

Klíčová slova v češtině

jazykové modelování, konečné automaty, rozpoznávání řeči

BibTeX

@INPROCEEDINGS{IrcingP_2003_Fittingclass-based,
 author = {Ircing, P. and Psutka, J.},
 title = {Fitting class-based language models into weighted finite-state transducer framework},
 year = {2003},
 publisher = {ISCA},
 journal = {EUROSPEECH 2003 PROCEEDINGS},
 address = {Geneva},
 pages = {1873-1876},
 url = {http://www.kky.zcu.cz/en/publications/IrcingP_2003_Fittingclass-based},
}