LAUSR

Abstract This paper proposes the use of quantitative indicators to evaluate the comedic success of Japanese “Manzai” performances without using semantic processing or time sequence information. The validity of the proposed indicators was verified by predicting the rankings of the final rounds and decision matches of ten M1 Grand Prix, a national-level humor contest in Japan, using leave-one-out cross validation. The results demonstrate that the proposed indicators are able to predict the ranking of Manzai championships as the mean prediction precision was 0.58 (rank correlation) for final rounds, and 0.70 (champion prediction accuracy) for the decision matches.