Handling temporal information has been a main concern of areas such as natural language processing, planning, and knowledge representation. In scheduling dialogues, i.e., dialogues whose goal is setting up a meeting in th,is paper, temporal reference is central, and the relations between temporal expressions contain much of the ellipsis and anaphora t o be resolved.A s a consequence, filling in missing information contributes t o the understanding of the text. In this paper, we first identify the various binary relations present in this kind of dialogue, characterize them, and, finally, present a mechanism for processing with temporal relations that is based on wing the constraints found during the analysis.
IiitroductionTemporal information has proven to be a central issue in areas such as planning, scheduling, natural language processing, and knowledge representation. In scheduling dialogues, specially those directed to setting up meetings, temporal relations are primary because they contain much of the anaphora and ellipsis to be resolved. Filling in missing information due to either anaphora or ellipis in our context has two main purposes: t o support the translation of dialogues, and to provide updated information for further interpretation. In the second section of the paper we describe a number of different approaches to processing temporal information. We then describe the Artwork project, which provides the context within which temporal information is characterized in this work. In the fourth section we describe the characterization of the temporal relations in the scheduling dialogues used in the Artwork project and the mechanism for filling in missing information. In the last section, we draw some conclusions and suggest some directions for future work.
BackgroundThere are different approaches to dealing with temporal information. The approaches that we have considered for our work are Allen's Interval Algebra (IA) *This research has been funded by DoD under contract Number:MDA9070-93. [l] , a cognitive approach to temporal reasoning developed by Borillo et al. [5], and a logic for calculating temporal interpretations in discourse defined by Lascarides and Asher [9]. Many different temporal reasoning systems have been the result of research on temporal relations whkch vary in computational complexity and expressive power. IA [l] has been the basis for the implementation and evaluation of a number of systems [3], [7], [12:I, [14]addressing the representation of qualitative temporal information about the relationships between pairs of intervals. According to IA, there are 13 basic relations (including inverses) that can hold between two intervals; e.g. 2 before y, its inverse 2 after y, and x e