Friend of mine did it via computational complexity: using gzip (as an approximation for KC) for attributing classical latin literature to their respective authors by checking which strings add the least additional complexity (due to shared writing styles, word choice, etc.) when compressed together and then clustering. Worked like a charm.

ETA: These were large bodies of text, however. Probably not gonna work for a bundle of comments, except for me, due to my overuse of "obviously", obviously.