Computational research and the case for taking humor seriously

Abstract

Abstract Computational research underscores the complex abilities underlying humor. Two decades of work have achieved substantial progress in some areas, notably systems that make jokes; detecting and generating laughter; and using irony in interactions. Sophisticated evaluations clarify both strengths and limitations. The achievements illuminate specific abilities, but also expose unsolved problems. The way humor pervades life is harder to match than self-contained episodes. Learning techniques are powerful, but providing the data they need is daunting. The medium is no longer simply verbal, but other modalities present deep challenges, such seeing the humor in a situation. There are real applications, but the most striking still depend on human support. Achievements and limitations together underscore the scale of the challenges involved in understanding humor.