Are the traits that contribute most to character scores "signatures" of some author's works?

[jwzimmer-zz]

Re #6 (comment)

• For this to be true, it would need to be the case that works by different authors have different top traits. This is verifiable.
• It would need to be the case that there exists some author for whom the traits were generally consistent across works. This is verifiable assuming such an author exists in the dataset -- just because no such author is in the dataset doesn't disprove it.
  • How would "generally consistent" be ascertained?
    • We could see how predictive a set of traits being important to a character is of that character being in a work by a certain author?
    • We could eyeball it?
    • We could see if characters top traits within an author's oeuvre are more similar to each other than to other characters, perhaps some kind of mutual information thing?
• If it is true, then do authors from the same literary school share similar "signatures"?
  • Do authors from the same time periods, same country, same culture, etc.?