The Read/WriteWeb team is a little
obsessed with next-generation search, even more so after Charles Knight's illuminating
post outlining 100
alternative search engines. Although Pixsta
didn't make that list (note that Charles is constantly updating the top 100), it is a
"visual browse and search" engine that has a unique approach to search. It caught my eye
today when Kris Hoet from Microsoft blogged it.
This is how Pixsta is described on its homepage:
"pixsta software reaches beyond the current text-based approach to search by automatically extracting visual content from images. we can organise large image collections into hyperlinked networks of visual similarity, so that users can browse the network to find images that come close to what they want, and also spark off new searches."
(emphasis ours)
London-based Pixsta's visual search technology is targeted mainly at e-retailers, and it aims to increase the all-important conversion rate of browsers into buyers. Whereas most search engines use text as their main variable, pixsta's algorithms are able to sort variables such as colour, outline shape and texture. In this respect, pixtsa is competing against Riya's like.com - which is also using image search technology for shopping. Also alarm:clock noted in a Nov 06 post another contender, Polar Rose.
]]> Kris' recent post points to two new developments at Pixsta, Elle’s browse&buy and Explore on Net-a-porter.com. On Elle, you click on an item and then a "click to see more like this" option becomes available:
Clicking on that gives you a range of similar items (in colour, style, etc). The point is that it doesn't just give you results by textual category ('black boots'), but uses other visual variables to give you hopefully a wider range of relevant results.
The Net-a-porter site is even easier, because you only need to mouse over a product to see its description and the 'click to explore similar items' option:
And when you click the explore link....
Conclusion
Visual and even 3D search is an exciting area, particularly for e-commerce sites. But you can imagine all kinds of scenarios where it might be more useful to search for visual clues rather than matching text - e.g. an online art gallery, entertainment, online video. It's great to see companies like Pixtsa and Riya exploring these new kinds of search technologies. Do you know of other similar efforts in visual or 3D search?
]]>