A controlled vocabulary for pictures may already exist
2.09 Humans depicted engaging in activities (xkcd.com)
One of the struggles of researching visual art online is that there is no controlled vocabulary to describe pictures. While it’s easy to do a Google Image Search on “stars,” it’s somewhat more difficult to get a computer to “see” the difference between five-pointed stars and four-pointed stars. Efforts to make this work are fairly cutting-edge and tend to float around the fields of Optical Mark Recognition (OMR) or augmented reality. Success in both areas is certainly possible but believed to be a long way off. Computers are great at math, but they have trouble with vision.
However, where academia fails, commerce often steps in. The US Patent and Trademark Office (USPTO) can’t wait for computers to learn how to see. Trademarks and logos need to be described now so the office can grant trade and commerce marks to companies. To do this the USPTO has developed a traditional controlled vocabulary for pictures. Your logo is a five-pointed star? That’s “01.01.03 A single star with five points.” You’ve scrapped the star and gone with a stylized image of the Big Dipper? The first category is “1.03 Constellations, starry sky” and the second is “01.03.01 Big Dipper, Little Dipper.” And so on.
The USPTO offers a guidebook with extensive visual examples. As a library cataloger familiar with book codes, this is the kind of thing that tweaks my geek. But aside from my personal interest, the potential applications for this type of code in the arts and digital humanities are significant. People would still have to manually enter the codes, but compared to the cost of developing an independent vocabulary or investing in augmented reality, the cost is easily justifiable. It doesn’t take a Ph.D. to recognize a five-pointed star.
Archive and museum collections that include USPTO could offer a better level of granularity in search and easier comparison between similar works of art.
