When Writing, How Anonymous Can You Be, Really? 184
An anonymous reader writes "Do you still think your online writing is, basically, anonymous? Think again! Research has it people put much of their personal traits into their writing, and computers may just be able to pick them up. That's at least what a recently announced competition on author identification (Given a document, who wrote it?) and author profiling (Given a document, what are its author's age and gender?) wants to find out. Alas, re-using other people's writing is no solution either; there's also a competition on plagiarism detection (Given a document, is it an original?). Wanna revisit your recent rants?"
Can it beat Google? (Score:5, Insightful)
I'd be interested to see if this program can do any better at analyzing my writing than Google does analyzing my search history.
Re:Can it beat Google? (Score:5, Insightful)
Google thinks I'm a 20 year old male. I'm in my early thirties and a gal. I think visiting Slashdot so much throws off its algorithm, as does all the video game sites I hang out at.
I think you misunderstand the purpose of the algorithm. A writing sample is, of course, insufficient to detect your age and gender precisely.
There is a good chance that your writing style matches that expected of a male in their twenties, in which case the algorithm had done well. You may be a gal, but your interests and behavior is perhaps more similar to that of a male in their twenties, and for the purposes of predicting what to sell you or what to expect from you, that's actually more accurate than your actual stats.
Authors can use these tools too. (Score:5, Insightful)
Of course, authors can use these tools too, and then iteratively change their texts until they cannot be correctly identified or profiled.
Just like spammers can check whether their e-mails ends up in spam filters before sending them.
It will be a never-ending cat and mouse game.
Re:astroturfers (Score:4, Insightful)
This would have been a lot more fun about two months ago to detect paid political astroturfers.
The ultimate AI-ish application would be an astroturfer plugin for chrome probably called "AstroturfBlock". So the site is a "tech" site, the contents are pure politics, and the text analysis system indicates an unemployed liberal arts degree holder... Go ahead and block it.
How is it going to detect whether people were paid to write something?
Re:That's pretty easy (Score:5, Insightful)
Actually, Tor comes prepackaged with a browser with privacy settings enabled by default. The server shouldn't be able to differentiate you from any other user of the stock Tor bundle.
That's for the TOR bundle if used as they recommend, but the article is about identifying authors by what they write, them not about idintifying by technical means. On Slashdot not RTFA could be used as an identifying metric but on the other hand it's a rather wide net.