Friday, January 27, 2006

Image searching

I recently read an article on image search. The search engine Ask Jeeves had launched its own proprietary tool for searching photographs and images on the web. One of the bosses of Ask Jeeves, Apostolos Gerasoulis, is pretty proud of the new tool as he is quoted: “Our new image search combines Ask Jeeves’ unique text ranking and clustering algorithms with sophisticated image recognition technologies and our new image index.”

It sounded very impressive, so I took a look. And of course I used the common searching routine: searching on your own name. It delivered 11 pictures and 1 image, including information on the URL. Of the 11 pictures you can see me on 10. One picture is of my Australian friend Louise, a photograph taken during our stay in Hong Kong. Was I impressed by the result of the search. No, not particularly. There were at least three duplicates (whenever are search engines going to develop a de-duplication algorithm Besides I know that there are more pictures on the web with a mention of my name; of course the whole search is based on text. A nice feature is the save routine of the pictures, neatly with a mention, that “pictures may be subject to copyright and you may need permission from owner to use a picture.”

Googles search result. I am on the front row, first one from the left side. I know that the photograph is copyrighted, but there is no mention anywhere, even on the site, where the picture was found.

While making inquiries, I though I might as well look at Google and Yahoo. Google showed 19 pictures, 18 photographs and one image. It was a wider selection than the one from Ask Jeeves; even a few pictures I never saw. Every picture has a mention of the URL, but also pixel information. But again duplicates, even one triplicate. No save routine and no disclaimer or copyright mention.

Up to Yahoo. Having typed in my name again, 4 photographs and three images appeared. One picture was there that I did not see on the other results. Of the three images only 1 made sense, the others ones were irrelevant. Bonus points: there was no duplication. On average however the result was very poor on quantity.

Am I impressed by these image search engines? No, not particularly. There are many more pictures of me on the web, but the engine did not find and index them (yet). Altogether, neither search engine demonstrated impressive results. I think that the developers will have to go back to their sweat cubicles and do better.

Tags: ;

No comments: