Understanding why and how a given answer to a query is generated from a deductive or relational database is fundamental to obtain justifications, assess trust, and detect dependencies on contradictions. Propagating provenance information is a major technique that evolved in the database literature to address the problem, using annotated relations with values from a semiring. The case of positive programs/relational algebra is well-understood but handling negation (or set difference in relational algebra) has not been addressed in its full generality or has deficiencies. The approach defined in this work provides full provenance information for logic programs under the least model, well-founded semantics and answer set semantics, and is related to the major existing notions of justifications for all these logic programming semantics.
