SRI à base d'inclusion graduelle

This paper investigates, from an experimental point of view, the use of graded inclusions, from fuzzy sets theory, to model an information retrieval system (IRS) as theoretically proposed by (Bosc et al., 2008b). Documents and queries are represented by fuzzy sets, which are paired with fuzzy operators. It is shown that the fuzzy logic settings are crucial in order to obtain a system suited for IR. With appropriate settings, it is possible to mimic classical systems, thus yielding results rivaling those of state-of-the-art systems, while preserving its strong theoretical framework. Conversely, examining operators providing poor results sheds light on the necessary properties of such a system. Last, the added-value of using this model is shown by considering possible extensions; in particular, a short experiment shows how one can make the most of morphological links in fuzzy IRSs. MOTS-CLÉS : modèles de SRI, logique floue, inclusion graduelle, implication floue, expressivité