“Google Image Labeler is a fun way to help us organize all the images on the Web!”
Simply. Google Image Labeler is a new game that Google has launched. You enter into competition with possibly millions of other people who have nothing better to do with their time than help Google do its own work by labeling images from its database.
It’s like back in high school when that popular kid came up to you asking for help, and you decide to take time out of your own busy schedule to do so. Sure, it’s a nice thing to do but this is a business, not some kid who might fail Algebra II if you don’t explain a few concepts.
For those unfamiliar with the Google Image Labeler interface, you are matched up with a partner and shown the same image. You then begin typing in descriptive terms for the image in hope that they match your partner’s terms, which are not visible to you. Once a term is matched, you are given 100 points and you move on to the next image. You have 90 seconds to complete as many images as possible. Passing an image is also allowed if your partner agrees to it.
I’ve played the game a few times and it is fun. It’s especially nice when you and a partner can match up pretty well, which isn’t often it seems. Other times, it’s hard to understand why someone wouldn’t describe it the way you are.
“Well, of course that is a tree. What else is it? Nature? Future paper products?”
Sadly, the majority of words I was matching with people were simple, one-word describers. I can’t imagine this being very good for an image description database. By using short words, people are able to move through more images and rack up a higher score. Has competition taken over the learning and helpful aspect? Don’t people care whether they’re labeling the images correctly or if they’re learning anything about the person they’re matching with? Just get the score up and thump your chest, please.
Thankfully, Google has a refining process for all of these simple descriptions. Google has designated “off-limits” terms that cannot be used by the current participants. These terms were already matched by a pair of labelers for that same image. By reusing images with these rules, Google forces users to become more descriptive if the first matched label was a simple one word term.
With the Image Labeler, it seems that Google has covered most of the loop holes, as usual. We’ll just have to get in there and have our fun to see if it will really be successful in the end.