Big data presents big opportunities — or at least it seems like it should, if only you can figure out the right questions to ask. For example, Yale librarians Lindsay King and Peter Leonard recently found themselves confronted with a century’s worth of Vogue issues that they’d gained access to through a subscription with ProQuest, which has digitized and tagged every issue of the eminent fashion magazine.
“It’s like you have a really great race car and most of the time it’s just sitting in the driveway,” King says. “What can you do with this much data?”
King and Leonard have found a few ways to start, and their efforts demonstrate the kinds of strategies researchers can employ with big data. They’ve used Bookworm — the tool developed by Northeastern historian Ben Schmidt (and applied recently to “The Simpsons”, as Brainiac reported last month) — to chart the frequency with which specific words have appeared in Vogue since its first issue in 1892. In this chart you can see that through the mid-20th century, Vogue spoke of “girls” more than it spoke of “women,” but by 1980, “girls” had all but disappeared from the magazine’s pages.
In many cases, though, the most interesting patterns aren’t ones you’d think to look for. “There are two methods. One is looking for something you think is there,” says Leonard, “and the other is letting data organize itself.” This latter kind of search is known as “topic modeling,” in which an algorithm creates constellations of words that appear near each other on the page. Using this method, King and Leonard noticed, for instance, that words related to “women’s health” cropped up a lot through the 1970s and 1980s, when the magazine was helmed by the apparently health-conscious Grace Mirabella.
Perhaps the most interesting pattern King and Leonard have detected, though, didn’t require a computer at all. Leonard had the idea to turn every Vogue cover into a transparency, and then to layer all the covers for a given year on top of each other. The results were surprising.
Earlier in the 20th century, covers were so varied and creative they just created an aggregate kaleidoscope of color — the magazine might feature anything from a glamorous portrait to a witty New Yorker-style illustration. More recently, they’ve featured celebrity models posed in a blur of different ways. But in the ’70s and ’80s, Leonard found, “There were so much similarity with the way women’s faces were displayed, you get this ghostly platonic image of a woman’s face.” To judge by their experiment, for the woman of 1980, there was one way to be — and it looked a lot like Brooke Shields.
Kevin Hartnett is a writer in South Carolina. He can be reached at firstname.lastname@example.org.