 |
Modeling Pronunciation Variation for Automatic Speech RecognitionRolduc, The Netherlands |
  |
|
Modeling Pronunciation Variation for Automatic Speech RecognitionRolduc, The Netherlands |
|
The present contribution has a double purpose. For one thing, it aims to introduce the pronunciation modelling - especially variability and context-dependency of the pronunciation - within the MHAT (Markovian Harmonic Adaptation and Transduction) framework. Lexical materials and a phonological are also defined within this framework. At the same time, within the above context, using spontaneous speech data through the telephone, three approaches of the pronunciation modelling for the continuous speech recognition are compared, the first one with a single pronunciation per word, the second one with context-free multiple variants per word and the third with context-dependent multiple variants per word.
Bibliographic reference. Perennou, Guy / Brieussel-Pousse, Laure (1998): "Phonological component in automatic speech recognition", In MPV-1998, 91-96.