Summaries of lubm100m dataset

Below we show the Weak, Strong, Typed Weak, Typed Strong, 1-forward and 1-forward-and-backward bisimulation summaries of a 100 million triples LUBM benchmark (saturated) graph.

How to read a visualization?

We can see that Weak summary is very compact. On the other hand, slightly larger Strong summary remains very informative and resembles E-R diagram. Typed summaries give an insight about the dataset structure as it was designed according to the benchmark. Finally, bisimulation summaries are much bigger and don’t give a meaningful picture of the data.

Comments are closed.