In which I have women's problems

[steepholm]

I was miffed by the GenderAnalyzer's conclusion that I was 99% male - based on certain 'strong indicators' that seem to be scattered through my LJ. Personally I can't see what these are - perhaps I'm the last to know - but I'm curious whether the analyzer works by looking at vocabulary alone, or collocations, or more nebulous things such as sentence structure. Or some combination of the three. Or, of course, something else entirely. (For what it's worth, it seemed to think quite a large proportion of my predominantly-female flist was male as well.)

Anyway, since it only looks at the latest page of one's blog in making its decisions, I should be able to sway it at least by a few percentage points, simply by including some vocabulary and phrases that seem more likely to appear in 'female' posts.

So, for the sake of testing the system, I'm dropping in the following:

My husband was in Miss Selfridge the other day, trying on make-up, when in came Victoria Beckham wearing a gorgeous ruched skirt. He's such a fangirl for David (squee!) that he didn't even notice. You know what men are like! If only I hadn't been at home with cramps, some knitting and the latest Stephanie Meyer I'd have been able ask for her autograph.

Results shortly...

ETA: Interesting - it wasn't swayed at all. Those strong indicators of maleness, whatever they are, seem to have overwhelmed everything else, and 99% likely to be a man I remain. Hmm.

Comments

[kalypso-v.livejournal.com]

I was 99% male, too (though I did eventually find one woman friend who was 100% female). I note that the poll on "Did GenderAnalyzer give the correct result for your blog?" is running at 52% correct, 48% incorrect. Without knowledge of likely gender bias in blogs, that sounds plausible for random results, doesn't it?

[steepholm.livejournal.com]

Yes it does - which is reassuring, I think. I wonder whether it might be the mention of Doctor Who that correlated so strongly with a Y-chromosome?

[rymenhild.livejournal.com]

I remember reading somewhere that the gender analyzer mainly looks for grammatical features. I don't remember which features matter, but I have a dreadful suspicion that formal prose with complex syntax is a sign of masculinity.

(If true, that's not problematic at all, is it?)

[kalypso-v.livejournal.com]

It sounds horribly likely.

[steepholm.livejournal.com]

If it works on grammatical features I'd be wondering whether those assumptions are pre-programmed, or whether it refines its algorithm as it goes, in the light of the results it gets. If - which I doubt - there really is a noticeable gender difference in writing styles it would more insidiously problematic than some programmer with bunch of stereotypes in his/her head.

[kalypso-v.livejournal.com]

I'm pretty sure that my "100% woman" friend mentions Doctor Who regularly!

It might be better on detecting sexual orientation, as I got something on the lines of "male at a guess but quite gender neutral" for a male gay friend, and "female at a guess but quite gender neutral" for a female gay friend. But that was a sample of two, so doesn't prove anything.

[kalypso-v.livejournal.com]

That's interesting... I've just been back, and now they think I'm 100% woman.

So either they've worked out a much better method for analysing gender in the intervening four and a half hours, or the results are entirely random. Unless they logged the fact that I'd used the poll to tell them they were wrong, which would give them 100% certainty about the correct answer?

I'd be interested to know if you got the same answer on a later visit.

[steepholm.livejournal.com]

I now seem to be 100% woman as well, by St Tiresias! And I didn't tell them whether they were right or wrong before, so that can't be the reason. I've tried a few other people, and everyone now seems to be a woman of some description, though not all are as extreme in their womanliness as thee and me.