A language interface to a polyphone-based speech synthesizer

Björn Gambäck, Martin Eineborg, Mikael Eriksson, Barbro Ekholm, Bertil Lyberg, Tomas Svensson

Abstract: The paper describes how the Swedish Core Language Engine, a general-purpose natural-language processing system was used as a preprocessing system for a polyphone-based speech synthesizer. The interface between the two systems utilizes the grammatical information obtained by the NLP system to improve the quality of the speech produced by the synthesizer. The interface is built in a modular fashion, with the first modules using the NLP information in order to assign sentence accent properly on phrase and word level, while the final modules decide on which entries from a polyphone library should be used and calculate the pitch pulse information used by the synthesizer.