Résumé en anglais

Workflow-level provenance declarations can improve the precision of coarse provenance traces by reducing the number of “false” dependencies (not every output of a step depends on every input). Conversely, fine-grained execution provenance can be used to improve the precision of input-output dependencies of workflow actors. We present a new logic-based approach for improving provenance precision by combining downward and upward inference, i.e., from workflows to traces and vice versa.