Situational aspects are very helpful to decide relevance but they have often been left aside by Information Retrieval models. The standard logical approach to Information Retrieval, based on deciding relevance with the entailment d j= q, where d and q are logical representations of a document and a query respectively, does not consider retrieval situations. In this work we propose to introduce these aspects through a revision process. The relevance test is generalized to be (S d) j= q, where is a Belief Revision operator and S is the logical representation of a retrieval situation. Besides, we analyze the basic properties of different belief change methods and their adequacy to the modelization of S d. We also study the close relation between our approach and other approaches dealing with counterfactuals.
David E. Losada, Alvaro Barreiro