I did something similar with books[1], but drawing the words based on their F-score was a brilliant move. I think it gives the viewer much more information than word frequency, and thus is much more appropriate for generating word clouds. I wish I'd thought of it first.
Dunno, I think the raw frequencies work fine in your case, because there aren't really any themes that all (or many) authors across your selections keep returning to.
But, thanks :-)
Yeah, probably. I'd have to do a hybrid weighted frequency/f-score thing, otherwise I'd lose some information, but yes, great idea with the F-score overall.