"Our model significantly outperforms a competitive baseline and generates funny jokes 16% of the time, compared to 33% for human-generated jokes."

From this paper:

Unsupervised joke generation from big data

Sasa Petrovic and David Matthews

The 51st Annual Meeting of the Association for Computational Linguistics - Short Papers (ACL Short Papers 2013) 
Sofia, Bulgaria, August 4-9, 2013

 


 

Abstract

Humor generation is a very hard problem. It is difficult to say exactly what makes a joke funny, and solving this problem algorithmically is assumed to require deep semantic understanding, as well as cultural and other contextual cues. We depart from previous work that tries to model this knowledge using ad-hoc manually created databases and labeled training examples. Instead we present a model that uses large amounts of unannotated data to generate I like my X like I like my Y, Z jokes, where X, Y, and Z are variables to be filled in. This is, to the best of our knowledge, the first fully unsupervised humor generation system. Our model significantly outperforms a competitive baseline and generates funny jokes 16% of the time, compared to 33% for human-generated jokes.

 

From The Register:

It uses 2,000,000 noun-adjective pairs of words to draw up jokes "with an element of surprise", something the creators claim is key to good comedy.

...

 jokes calculated by the software include:

  • I like my relationships like I like my source code... open

New to LessWrong?

New Comment
16 comments, sorted by Click to highlight new comments since: Today at 8:17 AM

Of course, the phrases that appear in the paper are the funnest...

Two comments, one trivial and one more important:

trivial: personally, I don't think any of the jokes are really very funny. Then again, I never really liked the particular formula they're using, and I'm willing to accept that my taste in humour might be a little idiosyncratic.

important: I might be stating the obvious here, but I don't think it's really that enlightening to compare this with human joke-making, as I think they're mostly doing different things. Although the computer can do a passable job at filling in the variables for a given formula, that's a far cry from coming up with the formula in the first place. That's the impressive act of creativity, and after-the-fact mimicry of the same is much less impressive. When a computer gag program is flexible enough to come up with less stereotyped, formulaic gags, then I'll be impressed.

I'm reminded of, for example, EMI, Experiments In Musical Intelligence, David Cope's computer program capable of writing startlingly good pieces in the style of dead composers - but, of course, unable to come up with a distinctive, expressive style of its own.

That's the impressive act of creativity, and after-the-fact mimicry of the same is much less impressive.

This reminds me of the following articles, which probably don't reflect your intended sentiment, but might help some readers:

Nice. For what it's worth, my post's sentiment was pretty much wholescale cribbed from one or two of the essays in Hofstadter's book, Metamagical Themas.

[-][anonymous]11y10

There's another abstract on this topic here: http://acl2013.org/site/short/2316.html

I'm not sure how to find the actual paper link, though someone posted this one below.

I haven't read the article or the paper, but in fairness, humans are also pretty good at finding humor in things (there are older examples).

I like my relationships like I like my source code... open

That's really good.

I've heard this joke before I heard of the paper, and found it funny. I'm surprised nobody's come up with that one before.

Humor generation is a very hard problem.

Unfortunately computers can't tell which of the jokes they generate are actually funny, which is an even harder problem. I wonder if there is any research in that direction.

EDIT: apparently they can.

Use joke evaluator to filter output of joke generator. Bam! Massive improvement on joke hit rate.