Discourse Representation Theory

Discourse Phenomena
Let's look at some discourse phenomena next.

Discourse Representation Structures
Do our previous observations mean that we have to give up our idea of systematically translating expressions of natural language into semantic representations? Well, yes and no. Yes, as we have seen before, using straightforward first-order logic gets us into all kinds of troubles. But no, there is a systematic way of translating if we use an intermediate level of semantic representation.A theory of discourse interpretation that uses such an intermediate level of representation is ^↗Discourse Representation Theory (DRT) , a formalism proposed in the early 1980s by Hans Kamp. As we will see, DRT deals with all the problems noted above (and more) in an interesting way. The invention of DRT led to a shift from a "static" to a "dynamic" view on natural language semantics. To explain certain discourse phenomena, such as the interaction between indefinite noun phrases and (anaphoric) pronouns in texts shown before, the traditional account of considering meaning in terms of truth conditions turned out to be unsatisfactory. DRT adopts the rather `dynamic' view of natural language semantics, where the meaning of a sentence is defined in how it can change the context.One of the striking features of DRT is that it, instead of working with first-order formula syntax, works with explicit semantic representations. Such a representations is called ^↗Discourse Representation Structure (DRS) , and describes the objects mentioned in a discourse and their properties. As we will see shortly, the DRT approach sheds a whole new light on discourse processing and has sophisticated means to deal with discourse anaphora.

Interpreting Discourse Representations
So far we have been concerned with how to construct DRSs for texts - but how do we interpret these representations? We will present three different ways for doing so:

• by virtue of the embedding semantics;
• by using context change semantics;
• by translating DRSs into first-order logic.

The first two approaches define an interpretation function with respect to models. The third method is an indirect approach - by translating DRSs into first-order formulas, we can just use the interpretation function for first-order logic (see » First-Order Logic).

Implementing DRT in Prolog
Now for the implementation.

Running DRT
Here an example call and the file listing.

Compositional Approaches to DRT
ABSTRACTDUMMY

Further Reading
Kamp's original formulation of DRT (Kamp 1981) is surely of historical importance, but for a thorough introduction, your best bet is Kamp \& Reyle's textbook (Kamp \& Reyle 1993). This voluminous book not only presents the basics of the theory, but also extends it with detailed coverage of quantification, tense and aspect, and plural noun phrases. For a complete different angle of approaching the theory, try Volume 2 of Gamut (Gamut 1991). You will find here a description of Groenendijk \& Stokhof's dynamic predicate logic and an alternative formulation of DRT. Another article worth mentioning here is Van der Sandt's ``Presupposition Projection as Anaphora Resolution'', demonstrating the virtues of explicit discourse representation structures to handle presuppositional expressions.Several compositional versions of DRT have been proposed, the earliest dating from the end of the eighties (Zeevat 1989). Since then several alternative proposals and extensions have been made. An interesting starting point for exploring compositional DRT is Muskens' work based on type theory (Muskens 1996). The framework of λ -DRT (Kohlhase et al. 1996, Kuschert 1999) is probably the most worked-out theory for compositional versions of DRT.