Making collaborative (simple) descriptions of digital objects by way of tagging has been a hot topic in recent years. Think of the retrieval tools offered at popular sites such as flickr, Last.fm, YouTube, or Amapedia. But works of art and other artifacts can be tagged as well. The databases that grow from such tagging behavior may actually mirror hidden visual knowledge, and thus become a source of more semantically rich (complex) descriptions.
The principle is simple: ask your audience to assign words or other symbols to a set of objects. Record the results of the poll in a database, analyze the data and especially: link and combine the records of different users, and feed the outcomes back to the audience. Does that sound listless? Well, tagging need not be dull. Google, for instance, has a fascinating Image Labeler, where the challenge is to label a maximum number of images in agreement with an (anonymous) partner. The winners of the day (i.e. the winning couple of image labelers) has assigned the highest number of same descriptors to the randomly presented images. In playing the game the gamer is actually assigning verbal labels to the images. Of course such labels can be used by Google in image retrieval operations. Once a bouquet of sunflowers in a vase, is always a bouquet of sunflowers in a vase.
In Google Image Labeler the labels are often describing subject matter, but if an image is formally biased in a distinctive way (very yellow, very symmetrical, etc.) labels denoting formal image characteristics may occur as well.
You can get an idea of serious research into tagging museum objects at Steve.museum, a project initiated by Jennifer Trant from Archives & Museum Informatics. The project considers the benefits (and possible limitations) of verbally tagging museum artifacts.
But visual tagging may be profitable as well. In 2000 I developped a small voting system for labeling works of art via Internet: Art.Similarities. A reduced version is online at: http://www.let.leidenuniv.nl/arthis/projects/art.similarities/. Students tested the application in the years 2003-2005 and I am preparing a thesis on visually denoting formal features of artifacts. In Art.Similarities observers are invited to link simple “attenuated” images (icons) to “replete” works of art. [Goodman, 1976] It appears that a specific kind of information/knowledge about visual objects (similarity of works of art), observers (visual behavior), and patterns of attenuated images (indexing potential) may come to the surface via an application like Art.Similarities.
Rense Nauta suggested to display just the tagging results whereupon the observer may try to visually figure out what these attenuated images stand for. See the result in this “icon cloud” viewer. It consists of only two screens, one displaying randomly selected visual labels, assigned to one particular item from a small collection of artifacts. The other – just click the arrow in the upper right hand corner – gives away the identity of the artifact under consideration. Und so weiter.