The semantic cohesion between utterances focusing on the presuppositions

Tadahiro KITAHASHI, Noboru BABAGUCHI, Seiichiro DAN, Atsuko TAKANO

The Institute of Science and Industrial Research, Osaka University

Ibaraki-shi, Osaka 567, Japan

e-mail: kitahasi@am.sanken.osaka-u.ac.jp

This paper discusses an overall model of the structure of "question and response" pairs which can cope with diversified patterns of utterance pairs and shows that the model is effective in dialogue understanding. The model is constructed on the basis of the following two points. One is that a hear's reply could be affected by speaker's misunderstanding of hear's situations and the other is that a reply to a question might convey some information of background related to a direct answer instead of a direct answer itself. In both situations, the presuppositions underlying the speaker's question play a crucial role in generating the reply. The patterns of the reply stemming from the cases of misunderstanding can be classified by analyzing the presuppositions without which the question does not make any sense. In such cases, the replies either point out the misunderstandings or search replies which satisfy the intentions of them. For example, consider the utterance u1 and u2 at a travel agency. u1: How much is the map in Chinese? u2: We don't have Chinese maps here. (a direct answer) There also could be indirect answers rhetorically related to the direct answer. Here the term "rhetorically" relation means a relationship between a utterance pair expressing "reason", "evidence", "elaboration", "contrast", "generalization" or "result". Starting with these two points, we construct a structure model of the "question-reply" relationships, which is a model of the internal process of human dialogue between participants as well.

Keywords: cohesion, dialogue model, rhetorical relationship, presupposition