Second International Conference on Spoken Language Processing (ICSLP'92)

Banff, Alberta, Canada
October 13-16, 1992

Prosody as a Cue for Discourse Structure

Shin'ya Nakajima (1), James F. Allen (2)

(1) NTT Human Interface Laboratories, Kanagawa, Japan
(2) The University of Rochester, Rochester, NY, USA

This paper describes how well prosodic information correlates with the topic structure of discourse. To investigate this correlation systematically, first we introduce the notion of utterance unit which can be viewed as the basic unit in conversations. We then define four topic boundary classes: Topic Shifty Topic Continuation, Elaboration, and Speech-Act Continuation. The prosodic parameters- onset, first-peak, and final pitch frequencies- are measured at these topic boundaries to show how these prosodic parameters vary with the topic structure. Finally, we propose a schematic algorithm which identifies the topic boundaries via the prosodic parameters.

Full Paper

Bibliographic reference.  Nakajima, Shin'ya / Allen, James F. (1992): "Prosody as a cue for discourse structure", In ICSLP-1992, 425-428.