The new scientist article I got it from makes an attempt to explain it. (There's a link to that article in the OP somewhere too.)Frankster said:Link doesn't work for me for some reason :\
Shame, seems a lot of fun and would like to know the scientific rationale if there is one, behind this program.
That honestly doesn't surprise me. OK, so that there were other tests was not something I knew about personally, but the article I got this from points out that while the program they're using to analyse text is something they created themselves, the markers and indicators they were checking text for came from research papers that have existed for decades.Stasisesque said:Not only is this mostly inaccurate.
It's also not even remotely new.
http://bookblog.net/gender/genie.php
http://www.nytimes.com/2003/08/10/magazine/10WWLN.html?ex=1061784000&en=843e4c97d49a9f82&ei=5070
http://www.nytimes.com/2003/08/10/magazine/10wwln-test.html?ex=1168059600&en=a6ad778afcb6699a&ei=5070
http://u.cs.biu.ac.il/~koppel/papers/male-female-text-final.pdf
The stupid part is the writer for new scientist (who you'd think would be fairly careful with statements about supposed facts) not only seems to think this can work accurately, (and be useful in detecting fraud of some kind), but that the existing software is about 85% accurate...
(And from the tests I've tried so far it's failed at least 2/3 of the time...)