Recovering Explanations from Transformed Rule-Based Ontologies

A recent study in computer science tackles the issue of extracting explanations from modified rule-based ontologies. Datalog rules are frequently employed to create ontologies within Knowledge Graphs, and rule reasoners often enhance these ontologies by reformulating rules for greater efficiency. Although these alterations maintain the facts that are entailed, they fail to retain the derivation structure, meaning that a proof tree based on the rewritten rules does not easily provide an explanation using the original rules. The paper explores how to derive a proof using the original rules from one based on the rewritten rules, determining its computational complexity and pinpointing two relevant languages for defining proof transformations. This work is significant for logic in computer science and affects the transparency and explainability of reasoning systems in knowledge graphs.