Abstract de la publi numéro 13102

In an information retrieval (IR) context, users usually submit queries with few terms and no operators (e.g., and, or, +). However, most of users' information needs involve more expressiveness (e.g., `Potato famine in Ireland, but not in Cork'). Our work deals with this category of queries that may be processed by geographic IR (GIR) systems to parse digital libraries according to spatial, temporal and topical criteria. We propose a GIR framework that supports expressive queries and aggregates results of a multi-criteria search. We also make experiments to verify that this approach improves the effectiveness of such search engines (improvement of 27% for topical criteria only, and of 54% for spatial and temporal criteria).