I’ve just been hacking around with one of my pages, trying to tweak what Google Image Search thinks it represents. The page contains a number of short paragraphs, each with a header and a single picture. The problem is that Google cleverly mines the text surrounding each image for keywords, above and below, meaning that one painter’s work appears under an image search for a completely different painter’s name. Irritating, if you’re a painter.

(I’m not the only designer to have this difficulty. Searching on my own name gives in third, fourth and fifth places Victor Selivanov, Sandra Quickert, and Jouko Väänänen — each considered me by Google — then from the same site the real picture of me, but named Samson Abramsky — whose own picture doesn’t appear in the results at all. Irritating, if you’re an infinite game theorist.)

I eventually reworked the page into a table, since there’s another site using that technique that also returns images for the same query, but only the right ones. It’s irritating though, since my WordPress plugins don’t play nicely with tables which means I have to write the html by hand. And of course I’m still not sure that Google is going to do the right thing.

Anyone know a cleaner way of doing this? The problem is basically, how to indicate that a page is divided into logically unrelated areas, so that keywords in one area should not be taken to relate to images in another. (I’m assuming that Google simply skips over <div>s for this sort of thing, although if anyone knows different then that would be a solution. Same WP plugin complaints, but slightly lighter on the markup than tables.)</div>