suggestions from Dario Taraborelli

[jlcx]

@dartar made some interesting suggestions for Wikidata relationships to add to CauseGraph (cg_rels).

• author (P50)
• cites (P2860)
• published in (P1433)

[jlcx]

I have some concerns about "published in". It seems like people are using it in different ways, sometimes just as a synonym for "part of". There doesn't seem to be a reliable and straightforward way to sort those out.

[jlcx]

The P2860 graph alone is huge. Almost half a million nodes, which is manageable, but over 3 million edges, which is less so. Due to its size relative to everything else that's in CauseGraph so far, I think it should be a separate graph from the beginning. Even this graph will probably be an indecipherable hairball without getting rid of some of these edges (and/or nodes), and I'm not sure how many edges would have to be removed to make it manageable. One idea would be to remove less-connected nodes (or at least their citations), starting with in-degree 0 and/or total degree 1, and then checking the graph after each step.