Monthly Archives: June 2008

What happens in metafilter…

Posted on by
5

Inspired by this Language Log post discussing what Mark Liberman describes as “a nice negative specimen of the phrasal template ‘What happens in X stays in X‘” [see also this 2004 roundup from Tenser, said the Tensor], I’ve spent a little time searching the Metafilter archives for applications on the site of what I’ve found myself thinking of as “the Vegas snowclone”.

I came up with a healthy crop of results spanning from very late 2003 on up to the present, most of which fall roughly into one (or more, but I’m keeping it simple) of a few loose semantic and structural categories as seen below.
Continue reading →

Days numbered in months

Posted on by
Reply

Odd phrasing from Josh Marshall over at talkingpointsmemo:

What does seem clear to me is that Lieberman’s days in the Democratic caucus, or more specifically, his days with a committee chairmanship courtesy of the Democratic caucus are numbered in months.

It’s easy to see how it could happen — Marshall wanted to convey the scope of time explicitly (e.g. Lieberman is in trouble come Election Day or thereabouts), and “Your months are numbered” isn’t any good — but adding units on to the end of the fixed phrase just doesn’t work for me.

Or it does work for me in a making-me-chuckle sort of way, I guess.

Announcing Garkov

Posted on by
4

I’ve just finished up the initial version of a new project:

Garkov — Garfield + Markov chains.

It’s a probability driven strip generator that builds Markov tables from transcripts of existing strips and then synthesizes and renders novel text as a new strip. The results are sometimes pretty apt and sometimes deeply weird, and I’m pretty happy with it so far.

Examples: Speak – The suspense – Crazy about you

Right now it’s working from about 500 transcribed strips — just a fraction of the 10,000 that have been published by Jim Davis over the last thirty years — and is using nine different background strips. Both of those numbers will likely grow as I have the chance to do more transcribing and strip-preperation.

In principle, this can be applied just as well to other comic strips. In practice, the code should make it reasonably easy to do so as well, so I hope to at least test another strip sometime in the not too distant future.

[Update 6/9/08 - response to Garkov has been pretty great, enough so that I ended rewriting the db code underneath it on Saturday after a bunch of traffic crushed the will out of my original rinky-dink Perl DBM solution.

Also, I update the link up there to actually go to the site. Sheesh. Sorry about that.]