Hi!
I could do the same for the whole Wikidata, instead of only the elements connected to Wikipedia, but to GerardM's point, why is this so bad? I'm
Well, I'm not sure about "bad", but certainly as you can see it won't be complete. Wikidata is a separate data source from Wikipedia, and while there is a huge overlap in the data, they are not identical. Moreover, English Wikipedia is not the only Wikipedia, of course, so limiting the lookup to English only would again omit data. Sometimes this way is good enough, sometimes it would not be.
Something I need to figure out now: How to navigate the tree of sub-classes. Probably I'll create a 'convolution table', but maybe there's something out there I can use?
This tree is huge :) And it may be also a bit weird. I can't remember from the top of my head any place that had the full tree... There's this tool: http://tools.wmflabs.org/wikidata-todo/tree.html but I think if you ask it to generate full tree it may not work well.