The site had orange headers, a three column layout with two vertical thin black lines, no content below the fold, top and bottom navigation, and a logo in the upper left corner. I used this site early in the week, and several days later I only vaguely remembered what the site was about. The site did not rank well enough on a short-tailed keyword for a successful search. The week was research heavy, resulting in an extensive history. At the time I did not expect to need the site ever again, so it was not bookmarked.
As I tried different search methods I became frustrated that I could remember exactly what the site looked like, but could not find it anywhere. A visual memory search engine was exactly what I needed.
The site like.com or ViPR for the iPhone are visual recognition search resources using existing images from pictures or a website you are visiting as opposed to a mental picture. How would using a mental picture as a keyword work? Current website code would allow for indexing a website based on its layout, colors, and images. The keyword entry could be a written description of the site, or a ‘draw what you’re looking for’ area. A database could then return ranked pages on the closest visual match.
When might we see this type of engine? Let’s look at Searchme.com. They present a pictorial results ‘page’, much like the possible results for a true visual recognition search engine. Searchme began in 2005; three years later they’ve got a usable product with more work to do. However, Searchme does not use Google or Yahoo-based results. I don’t expect the demand for a visual memory search engine to be high enough to validate a serious effort; however, if a preexisting database were used with a preexisting interface, I would expect a beta version well within three years.
The uses of such a search engine are more novelty than a necessary type of search. Is it the end of the world that I never found that site again? No. Would I have liked to? Of course. Beyond being useful for very visual searches, the engine could be useful for designers and programmers looking for ideas and solutions for their own websites.
Twenty-percent time Googler project, anyone?
