Přejít k obsahu


Fitting class-based language models into weighted finite-state transducer framework

Citace: [] IRCING, P., PSUTKA, J. Fitting class-based language models into weighted finite-state transducer framework. Eurospeech, 2003, roč. 2003, č. 1, s. 1873-1876. ISSN: 1018-4074
Druh: ČLÁNEK
Jazyk publikace: eng
Anglický název: Fitting class-based language models into weighted finite-state transducer framework
Rok vydání: 2003
Místo konání: Geneva
Název zdroje: ISCA
Autoři: Pavel Ircing , Josef Psutka
Abstrakt CZ: Článek pojednává o jazykovém modelování, které je založeno na třídách a je přizpůsobeno využití v síti transducerů.
Abstrakt EN: 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.
Klíčová slova

Zpět

Patička