numblr

Garkov is what happens when you combine Garfield comic strips and a Markov chain. The applications takes transcripts of Garfield cartoons, then generates new strips using automatically-generated text that should, in theory, be at least semi-coherent. But as with most applications of Markov models designed to produce coherent text: mostly, it is not, but it does have the potential for humorous results.

According to the blog post announcing it, it’s only using 500 transcribed scripts and nine background strips, but that number will grow (theoretically increasing the coherence of the resulting strips).

Regardless of how good or bad it is, people doing clever things with code is great.

(via hthth)