ITRW on Adaptation Methods for Speech Recognition August 29-30, 2001 Sophia Antipolis, France

Generating Non-Native Pronunciation Variants for Lexicon Adaptation

Silke Goronzy, Ralf Kompe, and Stefan Rapp

Sony International, Advanced Technology Center, Stuttgart, Germany

Traditional approaches to model pronunciation variations either require expert knowledge or extensive speech databases. In the cases where non-native speech is considered they are too costly, especially if a flexible modelling of various accents is desired. We propose to exclusively use native speech databases to derive non-native pronunciation variants. We use a phoneme recognizer to generate English pronunciations for German words and use these to train decision trees that are able to predict the respective English-accented variant from the German canonical transcription. In first experiments we achieved promising results using the enhanced dictionary for decoding accented-data.

Full Paper

Bibliographic reference.  Goronzy, Silke / Kompe, Ralf / Rapp, Stefan (2001): "Generating non-native pronunciation variants for lexicon adaptation", In Adaptation-2001, 143-146.