Arrange by subject
Every article goes to whichever of your topics it is closest to, and each topic
becomes an island whose volume is how much has been written about it. Islands you can
compare by eye; the bars below are the measurement.
A topic has no edge in a semantic space, so this sets one: an article counts as
about a topic when it is in the top 40%, 20% or 8% of the corpus for it. Watch the counts move —
that movement is the honest width of the answer.
loading the semantic space…