In this communication we devise a distributed modular scheme for organizing the different types of knowledge needed in the first phase of text-to-speech conversion, namely the conversion of the input text to a symbolic notation, representing phonetic transcription together with syntactic, semantic and pragmatic information. We argue that the proposed scheme will provide for easier formulation and maintenance of knowledge, and that its modularity will also simplify the implementation of different modalities, e.g. various reading modes.
Keywords: Text-to-Speech Conversion, TTS, speech synthesis, knowledge representation