Wednesday, February 11, 2004

Did He Really Say That? Or Did She?


Several weeks ago, Karen Lunsford asked members of the TechRhet discussion list for a link to an online gender checker. Paste some text into the Gender Genie, and the program will crunch the words around and tell you whether the author was a man or a woman.

Another list member, Elijah Wright, replied that the Gender Genie agreed with human analysis of the texts 85% of the time—as Elijah said, " pretty impressively good, considering how these things work."

Every computer program has its Achilles' Heel however. Elijah pointed out, for instance, that "The major failing [of the Gender Genie's analysis] seemed to be academic writing by women, which was almost always coded as "male" by the Gender Genie." (A point well-proven by the fact that I fed the preceding sentences through the Genie and found out that I am a man. I knew there was something wrong with me.) The site does include statistics on the success rate of the tool, according to reports from users.

What's perhaps more interesting than the analysis and stats, however, is the list of gendered keywords at the end of the analysis report. If only I'd known to use "with" more frequently...sigh. My downfall seems to be my extensive use of "the"—who would have known that "the" was a masculine keyword?

So what would I do with this tool in the classroom? First, there's a great opportunity to get students discussing syntax. Take a look at the linked background articles and talk about their discussion of how gender affects the structures people use when they write. And don't forget to explore the question of diction—just why is "the" coded as a masculine word anyway?

From looking at how the tool works, you can move on to trying the tool out. Students can submit their own text as well as the text of writers they've read. A trip to Project Gutenburg, the Electronic Text Center, or another collection of online books can provide your starting point. Grab some text, and see what the Genie thinks. There are plenty of interesting experiments possible:

—Does accuracy increase or decrease with older texts? Try comparing a Renaissance text to a contemporary text for instance.

—Is the tool accurate with both fiction and nonfiction by the same author?

—What affect does word order have? If you rearrange the words in the text, does it still score the same?

—Don't limit yourself to online books either—how does the Genie do on political and governmental documents?

—Does the Genie work as well on "spoken" text? Try feeding the transcribed text of an interview through and see how the Genie does.

—How does the Genie work on news stories? sports coverage?

—Is gender really an easy question? Does it make a difference if the author is gay, lesbian, or bisexual?

—Perhaps one of the more interesting options, can you change the text so that its meaning is not altered, but the Genie identifies the gender as changing? For instance, can you take a text the Genie identifies as by a man, change the phrasing, and trick the Genie into identifying the author as a woman?

Hmm. Just tried again. It still thinks I'm a man. I can't decide if that's a good thing or a bad thing?
tengrrl