ProvDIVE: PROV Derivation Inspection and Visual Exploration

Abstract

In a previous work, we presented a method to reconstruct PROV derivations from short social media messages. This method can capture a wide range of information spreading (and thus influence) among users, from explicit attribution like quoting to implicit means like content similarity. When applying this method to real-life datasets containing several million messages (e.g., a popular event), we are creating derivations in the same order of magnitude. To assess the provenance, it is useful to manually inspect the overall structure, the individual derivations and the users involved. Such tasks can be supported well by visualization techniques, yet thousands to millions of nodes are notoriously difficult to visualize.