Reasoning over linear probabilistic knowledge bases with priorities

Abstract

We consider the problem of reasoning over probabilistic knowledge bases with different priority levels. While we assume that the knowledge is consistent on each level, there can be inconsistencies between different levels. Examples arise naturally in hierarchical domains when general knowledge is overwritten with more specific information. We extend recent results on inconsistency-tolerant probabilistic reasoning to propose a solution for this problem.