Can a machine tell your gender just by reading your blog?
Posted at 11:09 AM Nov 06, 2008
By Bonnie Ruberg
A new piece of software created by Swedish programmer Jon Kagstrom claims to be able to tell the gender of a blog's writer at a mere glance. How does it work? Just type in a web address and the Gender Analyzer will tell you whether it thinks the blog in question was written by a man or a woman. The site learned to differentiate by reading 2,000 Blogspot blogs by known authors. From there it picked up words it classifies as male or female. For example, says Kagstrom, "beer" is 82.7% male, while "shopping" is "98.6 %" female. Here's how the Gender Analyzer classifies some of the site we know and love:
Heartless Doll: written by a woman.
NYTimes.com: written by a man.
Jezebel: written by a woman.
Salon's Broadsheet: written by a man.
Heroine Sheik (my own blog): written by a man.
As you can see, there are some problems right off the bat. 1) NYTimes.com is not written exclusively by men. 2) Broadsheet is written by women. 3) I am not a man. Says Kagstrom, "I think the reason for the somewhat low accuracy is because we have biased training data (only collected from blogspot)." Yeah, Blogspot isn't necessarily the height of cultural nuance, but it is interesting to see what stereotypes get dredged up by using blogs like that to determine how men and women write. Check out some example of how Gender Analyzer thinks the following words break down, and you'll see what I mean:
Technology: 99.3% male
Love: 89.9% female
Sex: 56.6% male, 43.4% female.
Well, at least some things are still nearly equal...
Comments
Worked for me.
Posted 11/06/2008 at 11:42:12 AMIt doesn't work with Nucleus blogs...
Posted 11/06/2008 at 12:43:45 PMAhem... BAHAHAHAHAHAHA
It thinks ToplessRobot is written by a woman. :D *dies*
Posted 11/06/2008 at 03:45:02 PMDoh! John you beat me to it!
Posted 11/07/2008 at 06:46:55 AMMy blog is written by a man. Won't my husband be surprised tonight in the bedroom?
Posted 11/07/2008 at 05:44:43 PMstuj
Posted 04/24/2009 at 10:57:58 AM