$AmbiPun$ : Generating Humorous Puns with Ambiguous Context

Anonymous

Abstract: Computational humor has garnered interest of the natural language processing community due to its wide applications to real world scenarios. One way to express humor is via the use of puns. A homographic pun plays on words that are spelled the same way but have different meanings. In this paper, we propose a simple yet effective way to generate pun sentences that does not require any pun sentences to train on. Our approach is inspired by humor theories that ambiguity comes from the context rather than the pun word itself. Given a pair of definitions of a pun word, our model first produces a list of related concepts through a reverse dictionary. We then utilize one-shot GPT3 to generate context words, and then generate punning sentences that incorporate context words from both worlds.We also investigate how the position of a pun word appearing in the sentence will influence the generated results. We compare our proposed $\textsc{AmbiPun}$ with well crafted baselines. Human evaluation shows that our method successfully generates pun 52% of the time, outperforming the the state-of-the-art model by a large margin.