Linguistic rules for text-to-speech synthesis

Prosody plays the role in speech ot producing a spoken message meaningful to a particular circumstance, from otherwise unweighted abstract units called phonemes. In the process of message formation, phonemes are given proper acoustic characterization according to the syntactic and semantic structure of the message, the location of stress and boundaries, and surrounding phonemes. Since this process occurs at a subconscious level, it has not received attention from researchers in the language related fields until recently. The recent progress in speech-synthesis techniques has made us aware of the importance of prosody in speech communication, and has encouraged the rule formation for this process. This paper concerns work at Bell Laboratories in the area of purely automatic (i.e., nonhuman) text synthesis. It summarizes the development of related prosody rules from natural speech data.