In a recent announcement at the World Wide Web Conference in Being, Google has come out with a splendid revelation. According to NYT, Google Research Team is currently testing a form of PageRank, which is designed specifically for images. It would be similar to what PageRank does for the web pages. This method will use a unique algorithm, that will take advantage of image-recognition softwares and fuse it with techniques for evaluating and ranking images. It is known as VisualRank
Even in this era of pictorial dictatorship, major search engines rank pages foremostly on their textual content and little stress is given upon the pictures. So far in the testing phase, VisualRank has produced higher quality results. Shumeet Baluja (Senior Staff Researcher, Google) said, “We wanted to incorporate all of the stuff that is happening in computer vision and put it in a Web framework.â€
For the testing phase of VisualRank, Google had focused on the 2000 most popular product searches such as iPod, Xbox and Zune. It then sorted 10 images from its ranking system as well as the standard Google Image Search. After that an image relevance scoring system was created. After the test, the researchers were able to successfully extract 83 percent less irrelevant images.
In the year 2006, an IT company based in Silicon Valley known as Riya, introduced Like.com. Its services enabled an user to go ahead with his choice of color, shape or any other visual attribute. Then the user was given the list of similar products available from other online companies. The fundamental of Like.com was that instead of focusing entirely on textual content, it focused on the symmetry of shapes and sizes.
Recommend this story
Navneet Kaushal, CEO PageTraffic is a trusted authority in the search engine marketing industry. He is a featured author at Web Pro News, Search Newz, Website Notes, DevWebPro, SEO Article and Web Help Now among many others.
@pagetraffic
{ 1 comment… read it below or add one }
well that would be great, coz images useful too in completing a whole content