There are lots of things in CompLing that I wish I’d thought of. Here’s one: the Gender Genie.

Just paste in some text (perhaps even a blog post from your humble Good Reason host), hit frappé, and it’ll give a guess as to the sex of the author.

I come up as male, as does PZ Myers, and (somewhat interestingly) Andrew Sullivan and TRex from Firedoglake. (Well, I had to check.) The wonderful Digby, had she not outed herself as female, would have been sniffed out by the Genie. Ann from feministing.com comes up male more than once. Walcott‘s running 50-50.

It works much differently than one might expect. It ignores all the manly or non-manly words like ‘explosion’ or ‘needlepoint’, and instead uses a very small set of function words. Each time you use ‘with’, that’s 52 points on the female side. Each use of ‘around’ gives you 42 male points. Women get pronouns: she, me, hers, we. Men get all the determiners: the, a, some, more. Add them up and see which side of the scale is heavier.

I don’t know what we can generalise from this about male/female communication, but it’s very cool.

I’m checking out Knol, a new project from Google. It’s a knowledge base that’s sort of a cross between Wikipedia and Amazon ratings. Here the focus is on authorship. Authors write their own Knols, instead of contributing to a group’s effort like some amorphous blob author. That means experts in a field can represent.

I can see how this would avoid some Wikipedia problems.

• Wiki-vandalism wouldn’t be a problem since you have control over your own Knols.
• No more pointless and frustrating edit wars on intractable topics.
• You don’t have to pretend to be neutral.
• Original primary research would be allowed, unlike on Wikipedia.
• It also avoids classification problems. Much discussion on Wikipedia concerns whether this article should be merged into that article. In Knol, the user does the classifying with only the search terms. A Knol could pop up under many different searches instead of just appearing in one article.

But is having so many separate authors a good way to arrange the world’s knowledge? One thing I’m finding in my language research is that individual points of view are terrific, if they can be aggregated into some kind of group opinion. Wikipedia does this by forcing people to hash out the issues and decide what content will appear on the page. Knol takes a rating approach, where individual votes from readers will (presumably) cause good articles to float to the top of the search results page. So it’d be like a bit like Wikipedia, except that there would be maybe 30 articles on generative grammar instead of one, and while each individual article might not be as strong as a good solid Wikipedia article, the best would come pretty close.

Here’s an example: I tried looking up ‘atheism’, and got one result. (Yeah, it’s early days.) It was a thoroughly useless page by “United Church of God” about how great God is, blah blah blah. Basically it was an old school web page that someone ported to Knol. But it’s getting downweighted mightily, and as more articles appear, it will probably sink out of sight.

The more I think about this, the more it grows on me. The next task will be some kind of true aggregation, where instead of reading only the best article about (say) the existence of UFO’s, you take the 100 top-rated articles on the topic, and automatically generate a group opinion. Sounds like a job for someone who’s into automatic summarisation and paraphrasing.

I bags it.