Production models as a structural basis for automatic speech recognition