MUS-Based Partitioning for Inconsistency Measures

Measuring inconsistency degrees of knowledge bases of different agents facilitates the understanding of an agent to her environment. Several semantic and syntactic based approaches have been proposed to quantify inconsistencies. By analyzing the limitations of existing approaches, we deeply explore the logical property of the Additivity condition through the connected components based on a graph representation of knowledge bases. Then, we show how the structure of this graph can be taken into account to discriminate in a fine-grained way the blame of each formula for the inconsistency. Finally, we extend our framework to provide an inconsistency measure for the whole base satisfying desired properties.