Effective Path Summary Visualization on Attributed Graphs

When a new dataset is modeled as an attributed graph or users are not familiar with the data, users may not know what can be retrieved from the attributed graph. Sometimes, users may have some intuition about the query, but how to exactly formulate queries (e.g. what attribute constraints to use) is still unclear to users. In this paper, we propose the idea of attributed path summary. In general, attributed path summary is a grouping of vertices such that vertices in each group contain paths from source to destination and the entropy of attributed values within a group is low and biased toward the intuition (i.e. preferred attribute values) given by users. We propose a novel 3-phrase approach which stitches key vertices together to form candidate paths and inflates those candidate paths into path summary. An extensive case study and experimental evaluation using the real Facebook graph that visualizes the path summary demonstrates the usefulness of our proposed attributed path summary as well as the superiority of our proposed techniques.