This is a model for Taiwanese speech recognition.