Jane Chandlee, Nanette Veilleux, University of Delaware
Research on discourse segmentation typically draws on the identification of certain cues in the various dimensions of text, speech, and gesture. Advances in automated segmentation models and algorithms have been achieved when these cues are taken into consideration. For gestures in particular, it must be observed that their presence and function as cues for discourse boundaries are both genre- and speaker-dependant. This study uses a recorded lecture to investigate whether speaker gestures can reliably predict the presence or absence of a discourse boundary and whether native speakers are able to make use of such a cue in isolation from others.