Stats for fileset dbpedia

Graph 1 shows the number of times nodes (or node pairs) of a given cardinality appear. So, if there are 200,000 nodes that appear as a Subject on three occasions, then 200,000 will be plotted at an x-position of 3 on the graph.

Graph 2 is more complex: it shows the cumulative entries to give a more readable graph. In this graph, if we have 100,000 nodes that appear as a Subject only once, and 100,000 nodes that appear as a Subject twice, then we plot points at (x=1,y=100,000), and (x=2,y=300,000). Thus, if a given Subject exists many times relative to the size of the dataset, it will cause a pronounced upward tick in the graph. This second graph is useful for showing the proportion of an index over S (or P, or SP, etc) that will be made up of small entries, vs large ones with repeating elements.

These graphs illustrate the number of times nodes are reused across all elements of a triple. Graph 1 shows the number of nodes that have been reused a given number of times: if 10 nodes appear 100 times, a point will be plotted at (x=100,y=10). Graph 2 is again more complex: if 10 nodes appear 100 times, and 2 nodes appear 101 times, points will be plotted at (x=100,y=1000), and (x=101,y=1202). This aids in visualising what proportion of the dataset is made up of heavily reused nodes vs rarely reused nodes.

Node Lengths

These graphs illustrate the length in bytes of nodes. In both cases, even if a Node is reused many times, it is only considered once in these graphs. Graph 1 shows the number of nodes that have a given length: if 10 nodes have a length of 100 bytes, a point will be plotted at (x=100,y=10). Graph 2 is again more complex, plotting the cumulative space used: if there are 10 nodes of length 100 bytes, and 2 nodes of length 110 bytes, points will be plotted at (x=100,y=1000), and (x=110,y=1220). This aids in visualising what proportion of space is taken up by nodes of a given size.