26 Brumaire CCXII (November 16, 2003)
The Gender Genie!
One of the things I do every weekend is check the Globe & Mail's columnists listing and make a special effort to read the ones by Rex Murphy and Leah McLaren (author of the infamous guide to being a functional alcoholic).
Anyways, today's piece by Ms. McLaren was on an algorithm that a group of scientists developed that analyses word content in prose and claims to detect the gender of the writer with an 80% accuracy level. It seems that a variety of newspaper columnists have sent their articles and their co-workers' articles through only to find out that all the female writers were identified as male. Of course, I'm not one to pass up a challenge like that, so I went through and selected long entries from a variety of different blogs.
So, I know what you want to know: what were the results? Well, I'm most definitely male. In fact, it managed to guess that four out of the five blogs submitted were male. Score one for it, or maybe not. You see, I'm definitely male, as is Derek, Matt, and Nancy (with an astonishing 0 for 5 record). As for the female… Well, sorry Peter, but it only thought one of your last 6 entries was written by a male.
Don't just take my word for it though, check it out yourself.
(And on a final note, this entry was also written by a male, although it seems to be solely my use overuse of the word 'the' that gives that side the advantage.)
Nancy and I played with this a while back. It's not just my last 6 entries that it thought were female, it's pretty much the entire blog.
However, if you look at how it analyses the data, you'll see that it follows a pretty simple rule. If you're talking about yourself, you're female. If you're talking about facts or other things, you're male. (This isn't me being sexist -- it's the algorithm).
The female newspaper reporters found that they wrote like males because newspaper articles tend to be about facts.
My blog says I'm female because I tend to talk about myself, and things that happened to me. The other blogs I checked at the time also turned out to be female.
I can't explain why it now thinks you guys are male, except that perhaps your blog entries have stopped talking about yourself as much.
It's because I rave about openGL. I dont say 'I' don't like it, I say its ass.
But its stiff funny that peter has turned out to be a drug dealer and a female all in one week. And I thought I knew you!!!
Oddly enough, I managed to find a link to the original implimentation of it. If you turn on the rule that interprets possessive pronouns, contractions, etc. (an option that didn't seem to get moved over to the linked implementation) then you show up as most definitely male.
I'll have to look for it again...










