This paper presents a computationally efficient and linguistically well-motivated methodology for providing Text-To-Speech (TTS) systems with all the information needed for constructing naturally pronounced and meaningful output. The method, which is especially suited to inflectionally rich languages as Greek and is presented in the context of the Greek TTS system developed at Wire Communications Laboratory independently and in the framework of Esprit Polyglot-I project, has greatly improved the quality of the output speech.
Keywords: Text-To-Speech synthesis, Prosody assignment, Linguistic analysis, Inflectionally rich languages