Friday, April 27, 2007

Sex machine

I had never tried Gender Genie, but Chris's* reminder was timely, since I recently got one of the periodic inquiries I receive from readers who wonder (given my somewhat unisex name) whether I'm male or female.

The answer is female, but is my writing somehow telegraphing testosterone? I asked Gender Genie, feeding it four Word columns. Sure enough, it says I sound like a guy: The columns scored, on average, male 1346, female 964.

As a blogger, I'm more girly: The item I tested was rated female, 840 to 521. But looking at the keywords the Genie scores, I suspect this was because of my quotes from The Economist's stylebook, along the lines of "Aggravate means make worse, not irritate"; those nots, for some reason, count as strongly feminine words.

But is it the writer or the topic that's being measured? Just for fun, I plugged in Barbara Wallraff's latest Word Court column. That scored masculine too, 1005 to 768.

So did Ruth Walker's last Verbal Energy column on the Christian Science Monitor website: male 1168, female 530. And linguist Heidi Harley's recent post on Language Log: male 713, female 495.

Turns out this is just what happened when a columnist at The Guardian put the Genie through its paces a few years ago: Only one of the newspaper's female columnists was identified (and just barely) as woman.

This does make me wonder: Is journalistic prose typically more "masculine" by the Genie's yardstick? And if so, what sort of prose was used to develop the algorithm for nonfiction? Some corpus, apparently, in which women use with, if, not, where, and be a lot more than men do.

*Corrected 4/29: I originally credited the Genie post to Josh. Sorry, Chris!

Posted by Jan Freeman at 06:17 PM

Print | E-mail to a friend | Permalink | Subscribe via rss