Word and syllable models for German text-to-speech synthesis

The correct pronunciation of unknown or novel words is one of the biggest challenges for text-to-speech systems. In this paper we describe the implementation of unknown word analysis as a central component of the text analysis module in the Bell Labs German text-to-speech system. The implementation is based on a model of the morphological structure of words and on the study of the productivity of word forming affixes. One important subcomponent of the word model is a phonotactic syllable model which enables the system to handle orthographic substrings that are unaccounted for by the explicitly listed morphemes. Finally, we discuss issues for future research. 1