An experiment in document retrieval using conceptual graphs

Abstract

In this paper an experiment using conceptual graphs to represent documents and queries is described. This paper is centred on the representation and answering language, a variant of simple CGs, and not on automatic indexing. Two important specificities of the experiment described in this paper are: first, it uses an existing general "ontology", the thesaurus of RAMEAU, which is used in all the french universitary libraries; secondly, the CG indexation has been obtained from the existing indexation by adding a new "relational" level. The main goal of this paper is to show that it is possible to add, to general traditional document retrieval systems, a CG level, and that it is worthwhile to pay an additional cost, because, using CG for indexing documents and for representing queries, leads to significant improvements in the answers obtained.