Abstract | The exchange of verbal and non-verbal com-
munication signals in face-to-face dialogue is
complexly organised in several ways: each
contribution is produced and processed incre-
mentally, contributions may be consecutive
(e.g. question-answer pairs) or overlapping
(e.g. backchannelling), and the contributions
themselves may be multimodal. Contributions
nonetheless exhibit pairwise utterance coher-
ence, and in two respects: across tiers and
across discourse co-texts. For these reasons,
we propose to distribute dialogue agents across
different tiers and to ‘incrementalize’ the se-
quential notion of turns according to the model
of music-inspired communication scores
|