Deanonymizer challenge

From /lounge/

A small userbase makes it easier to group posts together to form a pseudo identity, which may then eventually be deanonmyzed. But we are much harder targets than redditors and twitterers.

Sounds like a /prog/ challenge. First thoughts: build word frequencies for every post. Then loop over every post and every other post. For each pair of posts construct a link strength based on the word frequencies, then cull under a threshold, then the remaining networks represent identities (i.e. a human poster). You'd need some way of ensuring mutual exclusion too idk.

Useful terms to look up: Bayes, Markov.

Of course novel approaches preferred :)

Simple. I know my own posts, and it has been confirmed The Sussman has stayed with the dying ship on dis., so all other posts are yours.

>>3
What about me?

>>1
Cool. But how do we scrape this prague?

>>4
http://weitz.de/drakma/

Aaaand
https://github.com/gwkkwg/cl-html-parse

You want nodes with the class "postbody".

New subject: Write a new progscrape
Deadline: Before DistBB's final release.

You'll have to write a new one for the DistBB, though.

>>8
I may or may not have a go at that tonight.

>>8
Why don't we make DistBB with WebRTC?

>>10
Because I'd rather shoot myself in the cock than code in Javashit.

You will never figure out who I am.
El psy congroo.

>>11
Javashit is actually pretty nice when used well. Callbacks, closures in dictionaries, ect. You don't need to worry about parsing anything when you already have json. Asynchronous communication with callbacks is expressed pretty well.

>>13

You don't need to worry about parsing anything when you already have json

As a nitpick: http://timelessrepo.com/json-isnt-a-javascript-subset.

rererevvverseeee necroooooo