Second ESCA/IEEE Workshop on Speech Synthesis
September 12-15, 1994
With the integration of new services into the telecommunication networks, there is an increasing demand for automatic announcement and information systems employing oral communication. Since many telephone-network applications use large databases, synthetic speech is ideal for this purpose. One particular problem in speech synthesis is however, the pronunciation of proper names, because most tts-systems were initially developed to convert standard vocabulary. Moreover, names can be of very diverse etymological origins, their pronunciation often shows a significant deviation from the German standard letter-to-sound rules. Focusing on the demand of such new applications, we have developed a tts-system called PHRITTS in a joint project involving the Ruhr-Universiat Bochum and the Philips Kommunikations Industrie (PKIAG, Nürnberg, Germany). Since we are principally interested in good pronunciation and consequently producing highly intelligable names, we carried out a test to investigate how grapheme-to-phoneme conversion performs. This paper provides an overview of the TTS-system PHRITTS and reports the method and results of the evaluation test.
Bibliographic reference. Belhoula, Karim / Kugler, Marianne / Krüger, Regina / Rühl, Hans-Wilhelm (1994): "Evaluation of a TTS-system intended for the synthesis of names", In SSW2-1994, 211-214.