Context-awareness for information correction and reasoning in evidence theory

The notion of context awareness and the challenge of reasoning with partially reliable sources are two important aspects within Information Fusion. Context is the information relevant to, but not directly affecting, the problem at hand and can be broadly categorised into either context-for or context-of, referring either to the information related to some situation or to the environment induced by some situation, respectively. In evidence theory, the Behaviour-Based Correction (BBC) model generalises reasoning with partially reliable sources as well as contextual belief correction. In this paper, we propose a model for contextual reasoning framed into evidence theory, which captures both the notions of context-for and context-of. We rephrase the BBC model to explicitly account for variation of metaknowledge regarding source behaviour, and subsequently include within it the variables defining the context-for the problem and the context-for the source. The benefit is two-fold: on the one hand, the explicit inclusion of context in the reasoning provides a better insight into the problem and on the other hand, it can improve the expressiveness of the model. This is illustrated on a case of maritime surveillance involving a missing vessel, where it is shown that this model is not only more expressive than the simple fusion of sources model but also more interpretable.