» About     » Archive     » Submit     » Authors     » Search     » Random     » Specials     » Statistics     » Forum     » Facebook     » RSS Feed     Updates Daily

No. 2410: Another Garkov

First | Previous | 2015-12-24 | Next | Latest

Another Garkov

First | Previous | 2015-12-24 | Next | Latest

Strip by: WYE

Jon: Here, Garfield!
{Garfied starts gobbling up his food}
Garfield: All the mice. Get me a poker.
Jon: Okay, Tam.

Jon: I've comp bate high to my yout!
Jon: Is tifunnere's fring, a chom to youl gettle suit. My likend.
Jon: Yawarfield you ber ling ar le en.
{Garfield's words are partially cut off}
Garfield: Uh, justumb[...] t.

Garfield: Hmmmm.
Garfield: The Christmas decorations.
Garfield: It's here.
{Jon is hanging from the roof, wrapped in a string of Christmas lights}
Jon: My wall o.

The author writes:

Inspired by Josh Millard's Garkov, the results of which are featured in a couple of strips on here, I decided to create my own Markov-style text generator, and (using Josh's source text with permission) threw together a program that applied the results to Garfield strips.

The way my generator works is a little different from the original Garkov: instead of operating at the word level, it considers arbitrarily long "chunks" of characters. To generate text, it looks at the last n characters it's generated so far, appends a random character to follow it, then repeats the process. The probabilities are the same as in the source text: If n is 2, and, in the source text, "th" is followed twice by "e" and once by "a", then "e" will be twice as likely as "a" to be appended to "th" in the generated output.

It follows that the larger n is - the more characters are considered at once - the more coherent the output gets. The first and the last strip were generated with an n of 5, so they resemble the source text more or less closely, while for the middle strip I used an n of 2, and the result is more reminiscent of Finnegans Wake if anything. (Disclaimer: I have not actually read Finnegans Wake).

Some of the resulting strips (of which I've selected just three) are interesting for one reason or another, and I've found it quite fun to both program and play around with.

If you'd like to do the latter, the generator can be found at http://www.eric-kaiser.net/garkov.

Original strips: 1991-08-03, 2014-10-04, 2014-12-01.

[Ed: As with Garkov, future strips created using this new generator will not be accepted by SRoMG.]