Exploring Mainstream Natural Language Processing Techniques and Their Application to Computational Humor

Abstract

The purpose of this study is to document the creation of a software prototype that attempts to generate fact(s) about a variety of subjects in a legible English language format. Said prototype builds on top of an existing Natural Language Generation (NLG) research project called SimpleNLG. The format selected is that of the Harper’s Index, which is a list of facts in a rigid format published monthly by Harper’s Magazine. The rigidity of the format makes it easier to generate than regular language. This combined with the socio-economic and political topics that Harper’s authors usually touch upon makes it ideal for the study of Computational Humor, since it forces the reader to wrestle with their understanding of the world, a skill necessary to use and perceive humor. It is our hope that in creating this prototype, we can further the knowledge that goes into creating intelligent and adaptable computers that serve far more day-to-day applications in our increasingly technological world.