bing image graph bing

The Image Graph team gathered feedback and usage patterns for Bing’s Image search. From these, they determined that users mostly use Bing’s Image Search for shopping and for searching for recipes.

In this regard, the team aims to further enhance these users’ search experience by making it more satisfying and intuitive. They aim to reduce the amount of navigation that the user has to perform when running searches.

The Image team has found that about 59 percent of images on the Web have duplicates, which can range from one to thousands. These images are associated with their own queries and document texts. They also have very slight differences in their appearances, which make it difficult for users to find the exact image content that they need.

bing-duplicate-images-bing

 How Image Graph Works

To address this issue, Bing is designed to optimize the presented interface. It also includes the most relevant information online. In particular, Image Graph searches for and collects all duplicate images on the Web and sorts the information about them. It then groups the similar images and their metadata together to provide users with the most relevant data, in turn saving them time.

bing-image clusters-bing

Working with Microsoft Research has enabled the Bing team to build a more scalable and efficient graph model that has billions of nodes. It not only provides the right images to users; it also provides them with the most relevant descriptions of the images. When a user runs a search that references visually shopped for items, they will receive results that are labeled with shopping cart badges.

According to the Image Graph team, “These descriptions and links will help you learn more about a wide variety of subjects. The quality of the descriptions and the number of images with one will improve significantly over time.”

“Our goal is to help users be inspired, learn more and do more with image search. Our team mines through each of these sources to generate datasets in billions of content clusters to support each of these user goals. Examples include Best Representative Query (BRQ), Captions, Related Collections, Shopping Sources and More Sizes,” the Image team said.

With so many information available on the Web, users hardly have time to read them. As such, being able to base search results on images will certainly be a more pleasant experience and a real time-saver.