Dialogue material has been collected using bionic Wizard-of-Oz simulations of an automated voice-driven dialogue system; system turns, using rule-based speech synthesis, were systematically varied for both text and prosody. In a separate pilot experiment, subjects evaluated the system turns on a number of parameters. While they did not always clearly differentiate between the different conditions, they seemed to associate the "enhanced" prosody, preferred in an earlier laboratory experiment, with a more likeable personality.
Keywords: prosody, synthesis, dialogue, evaluation, simulation