Most Americans—some 55 percent—get their news from television. That’s more than double the number who look to the web first, and more than five times the number who turn to print.
So television news steers, to a great degree, our political discourse. But unlike text on the page or screen, it remains more difficult to analyze. TV arrives as sound and moving images—both of which algorithms have a harder time making sense of.
That’s why a recent pilot project from the Internet Archive is so welcome. Using the Archive’s massive archive of television news, Georgetown scholar Kalev Leetaru tracked all the locations mentioned on U.S. television news between June 2009 and October 2013, then plotted them on a world map.
On the foundation’s blog, archivist Roger Macdonald writes that the map constitutes the “first large-scale glimpses of the geography of American television news, beginning to reveal which areas receive outsized attention and which are neglected.”
Leetaru’s project isn’t the first to examine quantitatively how TV news represents the world. Last year, scholars in Germany and Israel examined how domestic TV news sources in different countries covered “foreign news.” MIT’s Media Lab, too, has mapped where the Boston Globe directs its attention; they’ve also looked at how often online news sources speak to men and women.
But the Internet Archive's map seems to be the first to depict the breadth of U.S. news. Looking at the map, I quickly found some rarely-covered areas that surprised me. Talking heads, apparently, rarely mentioned North Dakota, despite its booming economy.
To make the map, Leetaru used the massive library of closed captioning held by the Internet Archive. Macdonald explains the process, called “fulltext geocoding”:
These algorithms scan the closed captioning of each broadcast looking for any mention of a location anywhere in the world, disambiguate them using the surrounding discussion (Springfield, Illinois vs Springfield, Massachusetts), and ultimately map each location.
Underlying problems, therefore, might lurk in the data: “Two pairs of shoes” might be captioned—and thus interpreted—as “Two Paris of shoes.” But the prototype represents what’s now possible at the intersection of algorithmic text-reading and geographic visualization. Someday, data like this might inform more than the (worthy) world of media studies—it might help news organizations make decisions about where under-covered stories might be lurking.
One more thing: I’d love to see this map with the population filtered out: Does the I-95 corridor’s large population justify its massive coverage, or is it indeed over-represented? (It’s hard not to think of this XKCD classic.) I’m struck, though, by the visualization’s similarity to the famous Earth at Night image. It seems an image so awesome, so unfathomable, we can only always be referencing it in maps like these.