Before we could begin our attempt to model what made the strings funny, we needed to address a semantic complication that makes some funny NWs funnier than any purely formal incongruity measure could suggest: the problem of dirty words. Recognizing a rude word in a non-rude setting is notoriously funny. It was immediately obvious that some of the funniest NWs were funny because they were alluding to rude concepts, because the funniest-rated six NWs among the 5928 that were rated included the strings (which we present here without further comment) ‘whong’, ‘dongl’, ‘shart’, ‘focky’ and ‘clunt’.

Ok those are pretty funny. Of the “non-rude” ones (in appendix B), the top of the list actually does mostly feel funnier than the bottom, so eg. probble feels funnier than avelian.

Really neat application of Shannon entropy.