DRT is a powerful method for semantic representation and attempts to bridge the gap between syntax and semantics, which are probably two of the most important areas of research within Natural Language Processing (NLP). Relatively more progress has been made in syntactical analysis than its semantic counterpart. This is because semantics is still very much in the fields of cognitive psychology and philosophy, which makes it difficult to solve using the formal methods available to computer science. DRT, as its name suggests, is an overall theory of representing a discourse, and it is associated with Discourse Representation Structures (DRSs), which provide logical language-like features to DRT. DRSs are a type of data structure that are constructed using the Construction algorithm as described in Kamp and Reyle, 1993. Part of the Construction algorithm requires the use of a feature-based grammar to construct correct sentence parse trees and subsequently perform anaphoric resolution. The algorithm uses a set of DRS-construction rules which reduce the sentence parse tree and at the same time construct a DRS.
In this talk, I will present my implementation of a basic DRT system which uses a feature-based grammar to implement the DRS-construction algorithm. The system is designed in such a way as to allow one to easily extend its construction capabilities by adding new rules, hence giving the system the ability to processes more complex sentences. This extendibility is further enhanced by the object-oriented design of the system.
Edward Ivanovic is a postgrad student at the Department of Computer Science and Software Engineering of the University of Melbourne. He has a degree in Applied Science (majors in Computer Science and Software Engineering) and his current area of research is in modelling knowledge acquisition and semantic representation in a dialogue system.