Automated gender detection

There was a lovely piece by Alexander Chancellor in today’s Guardian about an algorithm that, when supplied with a sample of text, can predict (with 80% accuracy) the gender of the author. A simplified implementation of the algorithm is available on the Web. Chancellor found that the algorithm thinks that most of the female columnists on the Guardian are men. I tried it with two samples of my writing — one taken from my book about the history of the Net, the other from an email message to a friend. The algorithm concluded that the first had been written by a man, the second by a woman. Hmmm…