Fig 10 - Shakespeare in Context

This is not in the paper. I include it since it's a larger example, and includes a few subtle tweaks on the diagram design.

Here, we again look at the language of Shakespeare to see if the comedies can be "explained" by the kinds of words he uses (in the Docuscope sense). The difference is that we show his plays in the context of a large collection (about 550) of plays of the period. In this set, the 36 plays of Shakespeare are shown in more "intense" colors (e.g. dark and saturated green = Shakespeare comedies, lighter green = Non-Shakespeare comedies). This was an experiment to try to show two data fields (Shakespeare vs. Non, genre) with the colors.

From this diagram we can see that two features really tell comedies apart. A lack of "inclusiveness" (first person plural), and an excess of "Predicted Future." Part of the "comedy" of comedy is misunderstanding (so "we" doesn't occur), and setting up expectations.

This explainer was actually trained on just Shakespeare's plays, but it seems to apply pretty well across the larger corpus.

Large (630K) SVG explainer diagram (link in case it doesn't load correctly below)