Voice output for applications such as dialogue and information systems requires more than a plain or neutral tone of voice if it is to become more acceptable. These systems must be able to reproduce the acoustic characteristics of emotional and attitudinal effects driven by pragmatic procedures embedded within the dialogue controller. Possible approaches to modelling to this aspect of naturalness are discussed, and some sample data presented which illustrates some of the problems to be encountered.
Bibliographic reference. Morton, Katherine / Tatham, Mark (1995): "Pragmatic effects in speech synthesis", In EUROSPEECH-1995, 1819-1822.