They’ve made this into a “game.” When you start labeling you are paired with a partner (someone else who happens to be playing at the time). Each of you is shown the same series of images, one at a time, and you enter descriptive labels until the two of you match on a label. Points are awarded based on “how specific your label is.” At the end of two minutes you get a total score.
Now, I’m not sure how much this will improve image retrieval, because I often matched my partner on very simplistic labels like “woman,” “people,” and “black and white.” And sometimes I just had NO IDEA what an image was. so my descriptive labels may have been way off.
Try it for yourself, but I think the time limit influences players to stick with very simple labels and discourages coming up with more interesting (and useful) ones, which would be less likely to match your partner’s labels.
Believe it or not: This came to my attention at a meeting of librarians (Yes, that’s right, we get together sometimes, and you just never know what you’ll learn when we do!).