[jlcx]

For something related that takes a very different approach: https://causegraph.github.io/causalaxies

In contrast to the author's decisions here, I decided to

-go for an "everything tree" even if that will contain many more errors

-use DBpedia/Wikidata, and address issues discovered by editing Wikipedia/Wikidata

-use a 3D visualization tool, due to the size of the graph

I think it reveals an interesting overall structure, and some interesting details for those who zoom in despite the issues with the data.

[ricksunny]

OK this is amazeballs. Can you expound on technical implementation? Relying on Wikidata, then it relies on whatever Wikidata editors' internal governance system is for inclusion, exclusion, quality control of entries, correct?

I see you submitted this to HN almost a decade ago. How has it not gotten picked up for discussion??

[jlcx]

Thanks for the compliment! The larger graph is basically built by extracting all relationships of certain types (e.g. parent/child, teacher/student, cause/effect) from Wikidata, along with the earliest known date for the items (e.g. date of birth, time of invention/discovery). The layout started as a 3D force-directed layout, but I turned one axis into the timeline. For the visualization, I used (forked) https://github.com/anvaka/pm (which did get some deserved attention here: https://news.ycombinator.com/item?id=40817852 ) and the related ngraph.offline.layout repo.

Trying to answer your final question: there are a lot of things that I should probably improve here, but I've also wondered if this kind of giant graph visualization just doesn't really work for most people.