Provenance and Case-Based Reasoning

All Papers

Computational science takes a multidisciplinary approach to scientific investigation, tightly linking scientific research with computational studies and processes such as numerical simulation, data management, and visualization to study complex phenomena such as weather systems. The scientific importance of such processes has led to significant interest in recording the provenance of the data products, to support tasks such as assessing data quality. How to capture and use provenance information has become an important e-Science research area (Simmhan, Plale, and Gannon 2005). An interesting question is the possible relationships between this burgeoning provenance research and the capture and reuse of process information in artificial intelligence. This invited talk presents ongoing research exploring ramifications of provenance tracing and reuse for two different facets of case-based reasoning (Mantaras et al. 2005). First, it considers the opportunities presented by provenance capture as a source of process information from which to mine cases. Second, it considers the use of provenance tracking within case-based reasoning systems, to enable them to make better use of their cases. By mining the provenance traces of external processes, a case-based reasoning system can harness increasingly extensive libraries of process information to extend its case library. By tracing the internal provenance of its own cases, a case-based reasoning system can better manage and refine its own knowledge.