This paper gives an overview of the message generation component of a voice dialogue system. This takes a conceptual representation of the message, and generates a syntactically labelled surface structure using domain independent linguistic rules. At the syllable level, features representing prosodic focus are used to constrain sentence accent placement. The output is used to drive a speech synthesizer at phoneme level.
Bibliographic reference. Youd, N. J. / Fallside, Frank (1989): "Driving a speech synthesizer from conceptual input in the context of a voice dialogue system", In EUROSPEECH-1989, 1514-1517.