Towards Unified Provenance Granularities

Abstract

As Open Data becomes commonplace, methods are needed to integrate disparate data from a variety of sources. Although Linked Data design has promise for integrating world wide data, integrators of- ten struggle to provide appropriate transparency for their sources and transformations. Without this transparency, cautious consumers are un- likely to find enough information to allow them to trust third party content. While capturing provenance in RPI’s Linking Open Govern- ment Data project, we were faced with the common problem that only a portion of provenance that is captured is effectively used. Using our water quality portal’s use case as an example, we argue that one key to enabling provenance use is a better treatment of provenance gran- ularity. To address this challenge, we have designed an approach that supports deriving abstracted provenance from granular provenance in an open environment. We describe the approach, show how it addresses the naturally occurring unmet provenance needs in a family of applica- tions, and describe how the approach addresses similar problems in open provenance and open data environments.