Why Your Online Impersonation of a 16-year Old Girl Won't Last Long 137
An anonymous reader writes "Can computers pick up your age and gender from your tweets? If you want to give it a try, here's your chance: 'To develop your software for age and gender identification, we provide you with a training data set that consists of blog posts, Twitter tweets, social media texts, as well as hotel reviews.' Well, at least my paid Amazon reviews are safe for the time being..."
assuming too much (Score:5, Interesting)
I am sure you can pick up on general mood of a person, I am sure you can pick up various clues, but if somebody is set on hiding themselves by providing false information, I don't think you'll be able to identify them without what NSA calls 'meta data', as in pattern of your behaviour. I don't think it will be possible to do a better job than guessing at about 50/50 chance ratio about a person's age and gender if the person in question is actively trying to portray something he or she is not in a single conversation. From a pattern of behaviour? Yes. From a single conversation? ... this assumes too much about people. It assumes that 16 y.o. girls are also not pretending to be something they are not as well...
Tired of Slashdot? Come to comp.misc on Usenet! (Score:0, Interesting)
When a blatant attempt to plug a link such as this submission pisses you off, you have a choice. You can stick around and continue to be fed drivel, or you can come (back?) to Usenet where the air is clean. Eternal September is a good reliable free Usenet server, and comp.misc is the new official Slashdot replacement,
What about the following sites? (Score:4, Interesting)
The "dialectizer" http://www.rinkworks.com/diale... [rinkworks.com] "translates" English to Redneck, Jive, Cockney, Elmer Fudd, Swedish Chef, Moron, Pig Latin, or Hacker. And there's an English to Ebonics translator at http://joel.net/EBONICS/Transl... [joel.net] so it won't be that difficult to get a translator that outputs 16-year-old-girl talk.
Re:assuming too much (Score:4, Interesting)
Even if someone got a program that was "pretty good" at figuring out age and gender, I strongly doubt they could make it accurate enough not to be a huge civil liberties problem.
IF someone were to make a program that was 99% accurate, that would still be some 3million 16 year old girls in the US getting their houses raided in case they are a perv. (ok they aren't likely to raid on evidence like that alone, but they might use it as probable cause to 'dig deeper' into who she is... and if she caused the false flag then she is likely an outlier who the gov't wants to know more about anyway, for retraining... I'll go don my tinfoil hat now.)