SIGIR'98 demonstrations: Towards a Fast Precision-Oriented Image Retrieval System

Towards a Fast Precision-Oriented Image Retrieval System


Yves Chiaramella
CLIPS-IMAG, Université de Grenoble, BP 53, 38041, France.

Mourad Mechkour
CLIPS-IMAG, Université de Grenoble, BP 53, 38041, France.

Philippe Mulhem
CLIPS-IMAG, Université de Grenoble, BP 53, 38041, France.

Iadh Ounis
CLIPS-IMAG, Université de Grenoble, BP 53, 38041, France.

Marius Pasca
CLIPS-IMAG, Université de Grenoble, BP 53, 38041, France.


Abstract

The RELIEF image retrieval system is proposed as an integration of object-oriented modeling and Web technology, in an implementation of the logical information retrieval model on the O2 database management system. Following a relational indexing approach, it uses the conceptual graph formalism as the indexing language, and logic and standard information retrieval techniques to speed-up the retrieval process. The latter is performed by a polynomial matching function. Moreover, it includes relation-based inference which refines and extends the original matching function used in conceptual graphs, and improves the system performance in terms of recall/precision. Therefore, RELIEF is a first answer to the challenge of using expressive relational formalisms for accurate indexing in a precision-oriented system, while limiting retrieval time within reasonable values.


SIGIR'98
24-28 August 1998
Melbourne, Australia.
sigir98@cs.mu.oz.au.