Using Conceptual Graphs in a Multifaceted Logical Model for Information Retrieval

In 1986, van Rijsbergen promotes a principle in which the matching between a query and a document is founded on a form of uncertain logical inference. His formulation was deliberately abstract, no attempt was made to specify how such a principle might be implemented through an effective logic and under a particular semantics. In classical logics inference which is often associated with logical implication is too strict to be useful in the matching process. To overcome this problem, the possible worlds semantics seems to be a good way to formalize the uncertain inferences needed in an information retrieval system. Thus, in 1988 Nie initiated a work which has as its basic assumption the consideration that a document can be associated to a possible world. We think however that a more refined approach must be adopted especially when complex information are considered. It is based on the notion of facet, which allows to consider a document under several points of view, every one being associated to a possible world. Our claim here is that the conceptual graphs formalism is enough powerful to implement such a refined approach, in a way that captures the notion of uncertainty and that explains the retrieval process in terms of the possible worlds semantics.