figure out what a good visualization technique even is

[ACharbonneau]

Context: The model has multiple paths to any given search term and therefore often double counts entities. Karl came up with a partial solution to this "the stats table" that basically forces single counting by subsetting the data into groups of properties, like a 300 dimension venn diagram. The problem then becomes what to do with this data, because by subsetting in this way, it is no longer easily bar-plotable. The only plotting mechanism I know of for this kind of data is upset plots, but almost no one finds upset plots intuitive, and with the number of categories we have, the plots would come out huge and incomprehensible to basically everyone.

We basically need to find some solution that doesn't double count, but also is a plot that human brains can handle. The upset page https://upset.app/ has a few suggestions/examples

related issues:
#166
nih-cfde/dashboard#6