Looking at a few files, there's definitely some generated comments in there. Do you have any method to quantify how much of it is (likely) generated?