Role of the Context in Intuitive Combination of Evidence

Combination´ in `Theory of Evidence´ is merging evidence received from different sources to reduce the uncertainty one entertains regarding the problem in concern. Rules having different semantics and producing conflicting outputs are found in literature. Unfortunately proper guidelines are not available to select the combination rules suitable for a given problem. It has resulted in counter-intuitive use of the rules. Contextual information governs how the frames against which the evidence is expressed are related. It is however not embedded in the evidence and therefore not available to the agent applying the combination rule. This paper explores how contextual information can be extracted to be used for improving the intuitiveness of combination. It establishes a systematic way to extract contextual information, based on what the authors denote as hypothesis-level and frame-level inter-frame relationships.