ProZed: a multilingual prosody editor for speech synthesis

I present an overview of ProZed an aid for developing prosody rules for speech synthesis using the MOMEL and INTSINT algorithms and interfaced with the MBROLA, MBROLIGN and Praat programs. It allows the interactive editing of a symbolic representation of an utterance in any of the twenty languages and dialects for which an MBIPOLA diphone database is currently available. ProZed defines a number of different levels of representation of varying abstraction. At the lowest level, representations are a specification of the identity of each phonemic segment together with its prosodic characteristics. More abstract representations allow the user to abstract away from speaker-specific characteristics in order to concentrate on the meaningful content of the utterance´s prosody. It is evident that no tool is entirely innocent of theoretical bias and the tools described here are no exception to this rule. ProZed is, however, designed to be as theory-independent as possible so that it could be used to describe intonation patterns in a number of different frameworks with the ultimate aim of providing a framework that could be used to evaluate competing models of prosody. The program is currently implemented as a set of Perl scripts