An empirical account of the relation between discourse structure and pauses in Portuguese

The work reported in this paper is part of a project that aims at describing the role of prosody in conveying discourse structure in European Portuguese. We conducted an empirical study to investigate whether the organization of discourse could be reflected by the occurrence of pauses. In order to achieve that, we started by asking sixteen subjects to annotate two texts using two discourse segmentation methods: Grosz and Sidner’s, and Passonneau and Litman’s. We then confronted the segmentation obtained with pause distribution and length. We also verified which of the two methods can produce segmentation closer to pause occurrence. The results obtained are still provisional, but they at least show that the longest pauses are associated with the strongest discourse breaks.