Some Diierent Approaches to Drt

This paper discusses some issues in various approaches to Discourse Representation Theory. This is not intended to be detailed descriptions of various systems but general comparisons between both implementations of DRT and descriptions of DRT in some other context (e.g. a situation theoretic framework). The intention is to show how close (or how diierent) these approaches are and which aspects have been considered important by their developers. The view point is taken from a computational rather than a linguistic slant but these issues are not unrelated. 1 Background This paper discusses a number of approaches to Discourse Representation Theory , both in implementation and/or in description. DRT was originally described in Kamp 81] but has been developed in many directions. Although the original work was concerned with tense, DRT to many people has been inextricably linked with \donkey sentences", for which it gives good treatment. In this paper we will look at the following treatments of DRT: The \original" DRT Kamp 81] and its later deenitions in Kamp & Reyle 93]. Johnson & Klein 86] (J&K) which ooers a simple description of the core aspects of DRT and an implementation in Prolog. Pinkal 91] which concerns itself with a strict \compositional" approach to DRT. Cooper 93] ooers a description of DRT within the situation theoretic framework of EKN Barwise & Cooper 93]. Black 92] shows a treatment of DRT within the computational situation theoretic language of astl ((Black 93]) 1 | DRAFT | 2 Groenendijk & Stokhof 91b] describe the dynamic logic DPL which although is not a description of DRT itself is speciically designed to ooer a treatment of the same semantic phenomena as DRT and hence within this context seems relevant to our discussions. Also Black 93] gives DPL a treatment in astl and hence allows a closer comparison with DRT through the DRT description in astl. It must be emphasized that although we will give short descriptions of each of the above systems this paper is not designed as a full tutorial on these diierent approaches. Some knowledge of these systems will be assumed. What we will do is highlight certain aspects of these systems that allow us to give a closer comparison. Although we will try to make statements about what certain systems can and cannot do, ultimately this cannot always be determined merely from published papers (which typically are old). There may have been developments …