Gender Genie

I found a site called Gender Genie, that “uses a simplified version of an algorithm developed by Moshe Koppel, Bar-Ilan University in Israel, and Shlomo Argamon, Illinois Institute of Technology, to predict the gender of an author.” So I decided to do a controlled experiment. Taking publicly-available blog entries, it predicted that I am male, my brother is male, my brother-in-law is male, and Bob Mike is male.  Not bad, I thought.  So I tried the other set.  And the genie predicted that my wife is male, my sister-in-law is male, my brother-in-law’s girlfriend is male, and Bob Mike’s girlfriend Chelsea is male.  Not a bad algorithm, methinks, as long as you only feed it texts written by males.

Interesting thing is, I am nearly certain I could identify the sexes of each of these people by even a small sample of his or her writing. This implies that I’m doing something semantic (likely), some more sophisticated syntactic analysis (less likely), or both (oddly, I guess this is likely.)