Managing uncertainty in conceptual graph-based soft information fusion

The management and fusion of soft and uncertain information constitutes an important problem for decision support systems, yet it has been poorly studied until now. In this paper, we present an original approach, which addresses this problem. It is based on conceptual graphs, which is a formalism that allows the representation and fusion of soft information, and on a sound management of reliability inspired from a recent framework introduced in belief function theory. Our uncertain and soft information fusion process yields a probability distribution on a set of conceptual graphs representing our uncertain and soft knowledge about a situation of interest. Besides, we remark the formal similarity between such knowledge and a belief function, and we exploit it to propose a mechanism to compute a degree of certainty for any request regarding a situation of interest. Eventually, an example on maritime surveillance illustrates the capabilities of the proposed approach.