Un modèle Bayésien pour l'Agrégation des documents XML

In this paper, we are interested in aggregated search in structured XML documents. We present a structured information retrieval model based on the Bayesian networks theory. Relations query-terms and terms-elements are modeled through probability. In this model, the user’s query starts a process of propagation to recover the elements. Thus, instead of retrieving a list of elements that are likely to answer partially the user’s query, our objective is to build a virtual elements that contain relevant, non-redundant and complementary elements, that are likely to answer better the query that elements taken separatelly. We evaluated our approach using INEX 2005 collection and presented some empirical results for evaluating the impact of the aggregation approach. MOTS-CLÉS : recherche agrégée, réseaux bayésiens, redondance, complementairité.