Computational humor is the problem of generating, identifying and maybe even responding to humorous situations. It lies in the AI-complete category of problems, meaning that if it is solved, computers can be said to be as intelligent as humans. The seminar begins by summarizing existing theories of humor and their associated mathematical models. Then, we1 survey a paper that uses machine learning to classify text fragments as humorous or not, followed by a paper that algorithmically generates knock-knock jokes (you read it right!), and finally a paper that constructs pairs of graphs in order to generate conversational jokes.
Contact me at
<my first name> <my initials (hint: middle initial S)> at Google’s email service to reuse or adapt these slides.
1 This was presented as a seminar for an Artificial Intelligence course, in a team of three during my junior year. It was well received, with Professor Bhattacharya (now Director, IIT Patna) commenting, "Foundation, technique, recent advances - all aspects were dealt with with aplomb and insight. A model seminar, no doubt."