the tourist map of literature

[piranha]

marvelous and useful toy which enables one to find new writers one might like. i plugged in dorothy dunnett and it came up with some names i don't even recognise. exciting!

via matociquala.

Comments

[snippy]

Cool tool! Thanks.

[brooksmoses]

Hey, neat!

(Even if it is a bit offputting to have all the names wobbling around while they all figure out where to go on the map.)

[snippy]

Is it? I found it an appealing little dance. Wonder what it means.

[pir-anha.livejournal.com]

yeah, i liked it too; it looks so ... active. usually i can't stand it if things move (i hate flash advertising to the point where i'm ready to download a script to disable it), and animated lj userpics annoy me even if they're cute or funny.

[brooksmoses]

Perhaps I'm just being cynical about it -- and, really, my main complaint with it was that it was intriguing enough to watch that I got distracted from the actual content! (That, and it kept going a bit too long, I thought.)

What it looks like to me is this: imagine that all the names have springs connecting them to the other names. All those springs want to be some particular length -- some shorter, some longer, depending on the names involved. So if you had all the names in a tight little bunch in the middle of the page, and let them go, they'd all spring apart and then wobble around a bit as the springs pushed and pulled them into places where they were the right distance apart.

I think what that page (specifically, the Javascript program that's part of the page) is doing to place the names on the maps is a simulated version of that. It starts with all the names on the middle, and then for each name, it figures out what direction and how much all the various "springs" would be pulling, and then moves each name a little bit in the right direction, and then recalculates how much the "springs" would pull, and repeats this until they all finally settle down. So you get them all flying apart at the beginning, and then wiggling a bit into locations that are better and better fits to the distances between them that they "should" have.

It's a computational technique I've come across a number of times before for solving this sort of problem, but it's rare to see it presented in quite so visibly.

[kightp.livejournal.com]

Oh, my, that's cool! I just plugged in one of my favorite writers (Mark Helprin, as it happens) and found he maps to quite a few of my other favorite writers (Sherri Tepper, Ursula LeGuin, Robertson Davies), along with quite a few I've never heard of.

It would be fascinating to learn what variables they're using to connect people. Must bookmark this for later perusal...

[james-nicoll.livejournal.com]

Weird. It doesn't seem to have A. Bertram Chandler but it does have Tom Maddox, who wrote, AFAIK, one mid-list novel and two crappy X-Files episodes.

[pir-anha.livejournal.com]

that isweird. i am curious as to what database he draws on; i couldn't find anything about that.

maddox also wrote a few stories, but you're right, i wouldn't expect to see him in something like this. let me check something real quick... no, that's not it. i thought he might get the data from amazon.com, and that maybe chandler is mostly out of print and therefore has no "customers who bought this book also bought..." entries, and maddox might. but neither do, so that doesn't explain how maddox got in there. maybe he draws the data from several sources?

oh, wait, the guy is german. nope, amazon.de has the same results.

[daev.livejournal.com]

Whee! That's fun.

I was just reading that Kelly Link would be speaking at the Illini Union Bookstore, and then I go and type "Howard Waldrop" into this thing and her name is snuggled right up next to him. I don't know anything more about her, but I'm going to find out!

[prairierabbit.livejournal.com]

Very interesting, thanks! It connected some of my favorite authors whom I *thought* wrote very different sorts of things. Now for time to explore the new authors...

[pir-anha.livejournal.com]

i am pretty sure that those authors really do write very different sorts of things; i've noticed the same thing. i think it doesn't know anything about the authors other than that people who read A also read B and C and D, and it does some combinatorics with that.

if sufficiently many people who enjoy dunnett also like, say, cherryh, then cherryh ends up close to dunnett. on the surface this isn't a particularly insightful way of doing things, but it ties into the idea that folksonomies can sometimes be as useful as formal classification by experts.