SearchMe.com, the new visual search engine backed by Sequoia, has taken what seems like an obvious concept and built a search engine around it. Words take on totally different meanings depending on the context. If you type the keyword “comic” into a search engine, you could be intending to search for a comic book or a stand-up comic. So why doesn’t Yahoo or Google ask the end user for context clues? Seems like an obvious prompt, doesn’t it? SearchMe thought so. And now we have a search engine based on the idea.
Do a search for “comic” on SearchMe, and you’ll be able to pick amongst a slew of categories to narrow your search. As SearchMe increases in popularity and expands, so will its categories. The end goal is to have such a comprehensive list of categories that the user will be able to pigeonhole any conceivably confusing search term into the right category. This can certainly save the user heaps of time. The problem is that if the algorithm is worse than Google’s, the user won’t care.
But I’ve been fiddling around with SearchMe, and the algorithm seems pretty solid. I’ve not yet come across any search results that I felt were way off. Their algorithm appears to do a good job of categorizing sites properly. When you tell it to only show sites relating to comic books, instead of comics in general, the results are relevant. Must be a lot of latent semantic indexing at play.
Ironically, the worst feature of SearchMe is its visualization element. The website preview screen is too large and distracting. If they wanted to give the user a glimpse of what a site in the SERPS looks like, they should have just copied ASK’s binocular feature. Hovering over a result to see what a site looks like is more pleasing than a bulky window. If you could actually read the content of the website inside that bulky preview window, then SearchMe would have a hit a goldmine in terms of innovation, but you can’t because the text is too small. Frankly, it’s probably better that you can’t because publishers who sell ads on their site wouldn’t be particularly ecstatic about that feature.
SearchMe is relying on user feedback to improve their search results. This seems wise since a single person has the ability to catch mistakes that could affect millions of search queries down the line. I noticed a few flaws with certain keywords. A search for the keyword “bug” brings up ten appropriate categories, including insects, web development, and computer programming. But it doesn’t bring up a car company category, so people searching for the Volkswagen Bug are left out to dry. A search for the keyword “rover” correctly brings up the car company category, which will please Range Rover fans, but it doesn’t bring up the aeronautics category. If you were hoping to dig deeper into information about the Mars Rover, you’d have to do it without the special category tool.
These mistakes speak volumes about the impact of human intervention since it’s unlikely that SearchMe’s own algorithm would have ever caught them. If SearchMe’s popularity ever explodes, they can thank the beta testers. Making the feedback feature so prominent was a smart maneuver.