Humor in Collective Discourse: Unsupervised Funniness Detection in the New Yorker Cartoon Caption Contest

TL;DR

This paper evaluates various automatic methods for identifying the funniest captions in the New Yorker Cartoon Caption Contest, highlighting key linguistic features that correlate with humor.

Contribution

It introduces an empirical comparison of a dozen humor detection methods and identifies the most effective features for automatic funniness prediction.

Findings

Negative sentiment correlates with funniness

Human-centeredness and lexical centrality are strong indicators

Positive sentiment has a weaker correlation

Abstract

The New Yorker publishes a weekly captionless cartoon. More than 5,000 readers submit captions for it. The editors select three of them and ask the readers to pick the funniest one. We describe an experiment that compares a dozen automatic methods for selecting the funniest caption. We show that negative sentiment, human-centeredness, and lexical centrality most strongly match the funniest captions, followed by positive sentiment. These results are useful for understanding humor and also in the design of more engaging conversational agents in text and multimodal (vision+text) systems. As part of this work, a large set of cartoons and captions is being made available to the community.