Filling the Blanks (hint: plural noun) for Mad Libs Humor

Abstract

Computerized generation of humor is a notoriously difficult AI problem. We develop an algorithm called Libitum that helps humans generate humor in a Mad Lib R , which is a popular fill-in-the-blank game. The algorithm is based on a machine learned classifier that determines whether a potential fill-in word is funny in the context of the Mad Lib story. We use Amazon Mechanical Turk to create ground truth data and to judge humor for our classifier to mimic, and we make this data freely available. Our testing shows that Libitum successfully aids humans in filling in Mad Libs that are usually judged funnier than those filled in by humans with no computerized help. We go on to analyze why some words are better than others at making a Mad Lib funny.

“President Vows to Cut <Taxes> Hair”: Dataset and Analysis of Creative Text Editing for Humorous Headlines

“…We manually collected a set of 20 original news headlines and edited each of them such that some edits are funny and some are not. We asked several members of our research group to assess the funniness of each edited headline using the following integer scale developed by Hossain et al (2017):…”

“…Generating humor is a difficult problem. Past work includes Binsted et al (1997) producing punning riddles, funny acronyms from Stock and Strapparava (2003), jokes of the type "I like my coffee like I like my war, cold" by Petrović and Matthews (2013), and filling in Mad Libs R by Hossain et al (2017). Our headline work has the potential to help in humor generation, moving away from jokes with a strong template to more free form.…”

“…A big hindrance to progress on humor research is the scarcity of public datasets. Further- more, the existing datasets address specific humor templates, such as funny one-liners (Mihalcea and Strapparava, 2006) and filling in Mad Libs R (Hossain et al, 2017). Creating a humor corpus is non-trivial, however, because it requires (i) human annotation, and (ii) a clear definition of humor to achieve good inter-annotator agreement.…”

“Judge me by my size (noun), do you?” YodaLib: A Demographic-Aware Humor Generation Framework

“…The data annotation is undertaken by two parties: the players who fill-in the blanks to create funny stories, and the judges who assess the filled-in stories in terms of their funniness. We assume the threestage annotation framework devised by Hossain et al, (2017): judge selection, player selection and story annotation, with a few revisions to account for location-specific annotations.…”

A Neural Approach to Pun Generation

“…The inherent property of humor makes the pun generation task more challenging. Despite decades devoted to theories and algorithms for humor, computerized humor still lacks of creativity, sophistication of language, world knowledge, empathy and cognitive mechanisms compared to humans, which are extremely difficult to model (Hossain et al, 2017).…”