summercomfort | (no subject)

Stolen from

satyreyes: The Gender Genie counts how many times you use certain words and then decides whether the author is a male or a female: http://bookblog.net/gender/genie.php

So, some statistics, because I like that sort of thing:

My recent blog posts:
- 921 words, 959 girl points, 1457 boy points (0.66 girl)
- 608 words, 858 girl points, 994 boy points (0.86)

My recent longer LJ posts:
- 409 words, 270 girl points, 469 boy points (0.57)
- 396 words, 783 girl points, 287 boy points (2.73, or 0.36 boy)
- 385 words, 535 girl points, 585 boy points (0.91)
- 972 words, 1145 girl points, 1104 boy points (1.04, or 0.96 boy)

The intro to my BA: 1337 words, 900 girl points, 1880 boy points (0.47)
My Totalitarianism Unit Rationale: 1016 words, 652 girl points, 1350 boy points. (0.48)
Fic written 2001: 1078 words, 2007 girl points, 1342 boy points (1.49 or 0.67 boy)

Mussolini, Definition of Fascism: 919 words, 744 girl points, 1640 boy points.
Stalin, exerpt of 1930 report to Central Committee: 10965 words, 8658 girl points, 17193 boy points
FDR 1934 Fireside Chat: 2525 words, 1655 girl points, 3747 boy points

Neil's "How to Talk to Girls at Parties" (boy's perspective): 5207 words, 8001 girl points, 6546 boy points
Neil's "Snow, Glass, Apples" (female perspective): 5052 words, 6798 girl points, 6374 boy points

Conclusion: fiction = more girly?

Flat | Top-Level Comments Only

That's pretty neat. However, it's not so accurate, because out of my five or so most recent lj entries, it thought I was male for four of them, and it thought I was male for my paper on the typeface of the Gutenberg bible.

If I were a Feminist, I might say that because males tend to have higher social standing than females, Male Language is the Language of Authority, and so when we write nonfiction we try to write in Male Language. This makes it harder for a reader to distinguish the actual gender of the author.

But I'm not a Feminist; I'm more of a statistician. As a statistician I can't help observing that there are radial buttons for fiction, nonfiction, and blog below the text box. That means that assuming the program was intelligently compiled using multiple regression, it should have no systematic tendency to identify males as females or females as males within any of those three genres. That means that the "fiction = more girly" result is likely a statistical blip. It may be the case, however, that males and females write nonfiction in a more similar way than they write nonfiction (for whatever reason), resulting in lower accuracy. Sample size or consistency may also be involved.

So fiction does not = more girly, or at least these results don't directly support that conclusion. :)

Have you tried submitting the same text to all three buttons? I did the same text for both "fiction" and "blog entry", and got the same number, so I look upon those buttons with suspicion

Apparently when I write Lj entries I'm really female and when I write academic papers I'm really male. This somewhat shakes my faith in this theory.

well, academic papers are not fiction, right? Unless you think of BS as fiction... So non-fiction = more macho!

Interestingly enough, the same statistical algorithms they use to classify spam also work for classifying authors of given texts. My guess is that they just did something similar here. I also further assume that the text they trained the thing on were probably a bit skewed and possibly not as modern as they could have been.