Visualizations below show Weak, Strong, TypedWeak and TypedStrong summaries of DBLP dataset, available here. We can observe that summaries of this complex dataset can vary a lot. We weren’t able to plot bisimulation-based summaries as they were too large. We can also see that the Weak summary is collapsing the graph drastically into 2 nodes whereas the Strong delivers a meaningful picture of the data. Typed summaries benefit a lot in readability from the type generalization applied on the types hierarchy.
