LexDRESS - Speech Synthesis for a Speaking Pronunciation Dictionary - First Results

Authors: Rüdiger Hoffmann, Ursula Hirschfeld, Oliver Jokisch, Lutz C. Anders

Abstract:

The Institute of Speech Sciences and Phonetics at Halle University is developing a new dictionary of German pronunciation which will also incorporate a “talking dictionary”. This development is done in co-operation with the Laboratory of Acoustics and Speech Communication at Dresden University of Technology. A preliminary study published at ICPhS 2003 [1] showed that the quality of a standard TTS system (in this case the Dresden Speech Synthesis System DRESS) is not sufficient for the purpose of teaching the standard pronunciation. Therefore, the synthesis system lexDRESS was developed based on DRESS but showing features which are specifically important for a word synthesis from a close phonetic transcription. Using this close transcription requires a speech inventory which considers a large amount of allophonic sounds. A first version of the synthesis system as well as the inventory is available now. It is the aim of this paper to give an overview on lexDRESS. Furthermore, the results of a first evaluation are described here. A set of test words has been evaluated by experts and laypersons with regard to adherence and closeness to norms of pronunciation.