Thinking quite a lot about metadata just recently. Partly, it's that I've been uploading photos at such a furious rate there's just no hope of my providing much in the way of metadata for them despite flickr's relatively clean & simple interface for doing so. This is true of data proliferation these days pretty generally, I think. No matter how many small subsets have proper metadata attached, the larger ones' are never going to be that way, there are a lot of larger ones, the small, nicely structured ones get swamped, tiny island in a torrent.

There is of course, potentially useful metadata about my photos that isn't being used. For example, this text stream is being generated in parallel to the photo stream, both are by the same author, both have time stamps. There are corporate/standards issues preventing automated systems from recognizing the shared authorship (Google owns Blogger, Yahoo owns flickr, no standard cross platform method for establishing identity), despite the two streams being largely about one another, the only aboutness recognizable by automated systems is the hyperlinks from here to there. Without my tagging the photos, there's little hope of a machine recognizing the thematic relationships, for example, even if one could recognize the assimilation of street signs, warning signs, commercial signs to signs in general in the way that an attentive human reader of this could, it still couldn't recognize a photograph as being of a sign, though it might manage to recognize most of the text if it tried. The situation is even worse with all the animals and representations of animals.

Maybe robot iconologists are a lot to hope for, or something one shouldn't hope for- anyway, it's something I'm thinking about.