The fundamental function of an information retrieval system is to retrieve texts or documents from a database in response to a user’s request for information, such that the content of the retreived documents will be relevant to the user’s original information need. This is accomplished through matching the user’s information request against the texts in the database in order to estimate which texts are relevant. In this thesis I propose a method for using current natural language processing techniques for the construction of a text representation to be used in an information retrieval system. In order to support this proposal I have designed a matching algorithm specifically for performing the retrieval task of matching user queries against texts in a database, using the proposed text representation. Having designed this text representation and matching algorithm, I then constructed an experiment to investigate the effectiveness of the algorithm at matching phrases. This experiment involved the use of standard statistical methods to compare the phrase matching capabilities of the proposed matching algorithm to a sample of information retrieval users performing the same task. The results of this evaluation experiment allow me to comment first of all on the effectiveness of the phrase matching algorihtm that I have designed and more generally, on the usefulness of incorporating natural language processing techniques into information retrieval systems.