In other words, the paper’s authors consider funny 16% of random combinations of X, Y, and Z inserted into the pattern of “I like my X like l like my Y, Z”.
I actually quite like the idea of completely random gags in this formula. I feel like I might even prefer them to the lame ones given by the ostensibly smarter algorithm. It could be a rich vein of absurdist, non-sequitur laughs.
In other words, the paper’s authors consider funny 16% of random combinations of X, Y, and Z inserted into the pattern of “I like my X like l like my Y, Z”.
No, the completely random baseline generated funny jokes 3.7% of the time.
They get bonus points for their metrics of LOcal Log-likelihood (aka LOL-likelihood) and Rank OF Likelihood (aka ROFL).
They also get demerits for not discussing the error bars on their estimates given that they had only five testers.
I actually quite like the idea of completely random gags in this formula. I feel like I might even prefer them to the lame ones given by the ostensibly smarter algorithm. It could be a rich vein of absurdist, non-sequitur laughs.
I think that many nonsequitor jokes are funny more by targeted nonsequitorness than by randomness.