A case study on propagating and updating provenance information using the CIDOC CRM

Abstract

Provenance information of digital objects maintained by digital libraries and archives is crucial for authenticity assessment, reproducibility and accountability. Such information is commonly stored on metadata placed in various Metadata Repositories (MRs) or Knowledge Bases (KBs). Nevertheless, in various settings it is prohibitive to store the provenance of each digital object due to the high storage space requirements that are needed for having complete provenance. In this paper, we introduce provenance-based inference rules as a means to complete the provenance information, to reduce the amount of provenance information that has to be stored, and to ease quality control (e.g., corrections). Roughly, we show how provenance information can be propagated by identifying a number of basic inference rules over a core conceptual model for representing provenance. The propagation of provenance concerns fundamental modelling concepts such as actors, activities, events, devices and information objects, and their associations. However, since a MR/KB is not static but changes over time due to several factors, the question that arises is how we can satisfy update requests while still supporting the aforementioned inference rules. Towards this end, we elaborate on the specification of the required add/delete operations, consider two different semantics for deletion of information, and provide the corresponding update algorithms. Finally, we report extensive comparative results for different repository policies regarding the derivation of new knowledge, in datasets containing up to one million RDF triples. The results allow us to understand the tradeoffs related to the use of inference rules on storage space and performance of queries and updates.