Catalogue of Artificial Intelligence Techniques
Discourse Representation Theory
Categories: Natural Language
Author(s): Nicolas Nicolov
Discourse Representation Theory (DTR) treats the semantics and logic of natural languages. DRT represents larger linguistic units like discourses or texts and describes how syntactic form determines linguistic meaning. It provides a dynamic conception of meaning based on the observation that a human recipient of a discourse can process discourse on-line in an incremental fashion and the fact that new pieces of discourse are interpreted against the context established by the already processed discourse. Interpretation in DRT (i.e., the identification of meaning) proceeds in two stages. First, the incoming information is integrated sentence by sentence (incrementally) into the Discourse Representation Structure (DRS) for a discourse. By this integration process extra information, e.g., referential relations, may be added which does not form part of the compositional semantics of the individual sentences. Thus, a new piece of discourse updates the representation of the already processed discourse and the meaning of a linguistic expression consists both in its update potential and its truth-conditional import in the resulting representation. A completed DRS, i.e., a DRS which will not be further expanded, can be interpreted as a formula in Predicate Logic. Intuitively a DRS can be conceived as a partial model representing the information conveyed by a discourse. DRT provides a better treatment of anaphora resolution and quantifiers than Montague Grammar.
- Kamp, H. and Reyle, U., From Discourse to Logic
, Kluwer Academic Publishers, London, 1993.