PerryIt's a normal hobby to make graphviz visualizations of the categories of various pages on Wikipedia/Wikimedia Commons and then laugh at how absurd some of the inherited categories end up being, right? (I'll probably have some images later, right now I'm just seeing if I can get dot or even sfdp to build a ≈ 5000 node graph with labels.)
Time to open up Gephi...
Wow dot actually finished. I picked a bad article though; I was just curious to see if it'd do it.
Super Mario Bros at depth 4 is silly. Topmost colors are random but they "mix" as you go down (gvcolor). Edge colors are randomish and are mainly to distinguish crossings (edgepaint). See https://pprogs.blog/mario.svg for version with links so you can actually see what the categories are.
I didn't have any big luck with gephi and higher depth graphs (as in getting something interesting looking) but I did have some fun trying to set weights via Neo4j and minimum distance to the root node. Turns out that graphviz has `dijkstra` which does EXACTLY what I needed and it seems way faster (how??) than the Cypher query I had. I'll probably play some more with that tomorrow instead of doing anything important.
In the past when I've done this it was with Commons categories which tend to be much funnier and in some circumstances a bit more orderly I think? I dunno, a comparison would be fun.
This is all done via the amazingly awesome vcat as a base by the way. https://meta.wikimedia.org/wiki/User:Dapete/vCat#English
Maybe I should make my own fork of it that specifically automates this kind of visualization.
To illustrate what I mean regarding Gephi...
Basically I dunno how to do good hierarchal graphs in it. I'm inversely weighting by distance to the root node, but it still ends up being a giant cobweb.
And as an example of how silly Commons categories can be... "Actinostola.jpg" has a parent category of "Buildings by Eero Saarinen".
https://commons.wikimedia.org/wiki/File:Actinostola.jpg
Buildings by Eero Saarinen → University of Chicago Law School → Faculty of the University of Chicago Law School → Barack Obama → Barack Obama citizenship conspiracy theories → Donald Trump → Presidencies of Donald Trump → Second presidency of Donald Trump → 2025 United States government online resource removals → Files at risk due to the 2025 United States government online resource removals → PD US NOAA
BirderScott@perryprog six degrees of Actinostola bacon
@sfraser mmmmm anemone bacon
@perryprog fact: until today nobody IN MY ENTIRE LIFE has mentioned "anemone bacon"
Came up with this which gives some pleasing results. This time colors are assigned randomly starting at the file's categories and they then flow and mix upwards so you can see where categories start to dilute into one another. Dot size is proportional to degree (number of subcats + parent cats), with the black dot at the bottom having arbitrary size and being this starting file:
https://commons.wikimedia.org/wiki/File:The_Bug_Peek.jpg
Clickable version of this image is at https://pprogs.blog/svgs/bugpeek.svg
#Graphviz oneliner (save the downloaded .dot locally if you use this!)
DEPTH=5 WIKI="commonswiki" PAGE="File:The_Bug_Peek.jpg" curl "https://vcat.toolforge.org/render?wiki=$WIKI&title=$PAGE&depth=$DEPTH&limit=100000&format=gv&links=wiki" | dijkstra ROOT0 | gvpr -c 'BEG_G {concentrate="true"; mclimit=100; label=""; flow="back"; Defcolor="black"; nodesep=0.04; ranksep=0.1} E {arrowhead="none"; penwidth=0.2} N {shape="point"; width=degree/10.0} N [$.name == "ROOT0"] {width=3.0} N [$.dist == 1.0] {color=sprintf("%f+0.8+1", rand())}' | dot | gvcolor | dot -Tsvg -o out.svg