Skip to main content icon/video/no-internet
icon backReturn to Entries

Computational Humor

The term computational humor refers to an ability of a computational system to detect or generate some form of verbal humor. It does not refer to jokes about computers or any other computer-based systems. The textual forms of humor are typically not specified, but due to the complexity of the problem, they are often restricted to the low-hanging fruit, such as template-based light bulb or knock-knock jokes. To understand the nature of the phenomenon of computational humor, what needs to be discussed is the usefulness and applications of computational humor, as well as some of its dimensions, such as top-down versus bottom-up approaches and generation versus detection.

The usefulness and motivations of computational humor have been argued in many fora, starting from the 1990s, when the first papers appeared and the first workshops started to take place, to present days. From its inception, there have been two strands of arguments in its defense against those who thought it a doubly frivolous idea (humor in general can be seen by some as frivolous, topped by computing this frivolity): one based on promoting it as a method of advancing the knowledge about human humor competence, or humor theories, and the other based on the usefulness of computational humor applications.

Computational Humor as Method to Advance Theories

The computer is a perfect tool for testing an explicitly stated hypothesis: It can only do what it is programmed to do (provided that it is programmed correctly), without making any unspecified assumptions. What this means is that, in principle, a computer can be more reliable than a human in testing a theory: Its attention doesn't wander, it doesn't get fatigued, it doesn't forget—moreover, its knowledge can be usually traced to the source.

To test a humor theory on a set of jokes, a human or a machine needs to have access to the knowledge on which the joke is based. The details of this knowledge will vary from person to person, and it is very difficult to describe the extent of the knowledge that one has. Moreover, people may weigh various situations differently depending on what recently happened to them or on their personal circumstances. Again, this is very difficult to trace, but it affects the response to a joke. A computer, on the other hand, can always show the exact knowledge that it has used to process a joke and the weights that it has assigned depending on previously accessed information. Thus, it is possible to see all parameters that were in play for a computer for the analysis of a particular form of humor.

At the same time, for a computer to be able to analyze humor, very specific instructions should be given, as a machine cannot follow through a vague command, especially a command that counts on human intuitive interpretation to implement. Thus, a true theory must define its terms explicitly and formally to the smallest level of detail. Again, if these definitions are done correctly, and the system finds, for example, a text that it classifies as nonhumorous but a person finds it to be humorous, a counterexample for falsifying a theory has been found. This means that, in its current shape, the tested theory is not valid. Similarly, if a text is found to be humorous by a computer but not by a person, then modifications to the theory (model) should be made.

...

  • Loading...
locked icon

Sign in to access this content

Get a 30 day FREE TRIAL

  • Watch videos from a variety of sources bringing classroom topics to life
  • Read modern, diverse business cases
  • Explore hundreds of books and reference titles

Read next

More like this

Sage Recommends

We found other relevant content for you on other Sage platforms.

close

Have you created a personal profile? Login or create a profile so that you can save clips, playlists and searches

close
close
close Signed In to profile You are signed in as:
close
Logged in Institution Institution