images should be linked to senses not entries
When a lookup of a word occurs you display multiple definition senses and multiple images. Why not allow users to click on specific senses and then specific images to indicate which images go with which senses. An image is specific to a sense, not to a whole entry. Also, an image "suggests" a more detailed description of the word. E.g., I looked up "Ball" and one of the images was of a cat puffed up like a "ball of fur" --I would have liked to contribute to the further specification of why the images were related to various senses for ball, or related to phrases involving a ball. But the images didn't have anything that would have allowed me to comment, further link them (to senses, for example), etc. Think of images as parallel to senses and a match-up game between images and word senses....
-
Lee Davis-Thalbourne commented
The way the Flickr API works, I'm not sure how well that's likely to work. Flickr tags are words - that's all, really. There's no semantic information attacked to Flickr tags.
With that said, the Arachnophobia page's images being composed of spiders is seriously uncool for people who actually have arachnophobia.