A Knowledge-Based Approach to Robust Parsing

The research presented in this thesis describes the design, implementation and evaluation of a natural-language processing system that is used as part of an information retrieval system. Specifically, I focus on the development of a system that performs robust syntactic analysis of scientific texts written in English. The context of this research was provided by the Condorcet project. This project aimed at the development of structured concept systems to allow the use of structured concepts as indexing `terms' for effective and efficient retrieval. Moreover, it aimed at the development of a method to assign index concepts to machine-readable documents on the basis of semi-automatic analysis of title and abstract.