In this dissertation, a new logical model of information retrieval is developed and e￾valuated experimentally. This model is built on a general technique for uncertain rea￾soning called probabilistic argumentation systems (PAS), in which propositional logic
and probability theory are combined to represent and handle uncertain knowledge, both
in a symbolic and in a numerical way. The logical model respects the characteristics
desired for a logic of information retrieval, and interprets van Rijsbergen’s Logical Un￾certainty Principle in an original way. Propositional logic is shown to be a convenient
way to model information retrieval, at least when associated with probability theory in
the context of PAS