Google is moving to improve its search engine with new acquisitions and new test features that could help it ward off its emerging competitor Bing. On Friday, Google acquired Like.com, a startup search engine that combines computer vision and machine learning to deliver shopping search results.
Like.com's technology will let Google's search engine understand visually what terms like "red high-heeled pumps" and "floral patterned sleeveless dress" mean -- and even understand whether those pumps clash with the dress. Details of the acquisition were not disclosed.
"Since 2006, Like.com has been moving the frontiers of e-commerce forward one step at a time. We were the first to bring visual search to shopping, the first to build an automated cross-matching system for clothing, and more," said Munjal Shah, cofounder and CEO of Like.com. "We see joining Google as a way to supersize our vision."
Improving Google Shopping
Google may see Like.com as a way to combat Bing's Visual Search features. Bing Visual Search formulates and refines search queries through imagery, particularly for sets of results that tend to be more structured. Microsoft calls those data groupings galleries, and shopping is a key category within Bing Visual Search.
"This is about improving the shopping experience on Google Product Search and in mobile. Google has largely neglected its e-commerce site until very recently, yet it's still the largest shopping engine by traffic," said Greg Sterling, principal analyst at Sterling Market Intelligence.
As Sterling sees it, there remains a largely untapped opportunity for Google to dramatically improve e-commerce on its site. But more intriguing is how this might be used in mobile, perhaps in conjunction with Google Goggles -- the company's camera-based search tool.
"Like.com's visual database and capabilities will be folded into Google shopping and other products. That may give people the ability to take a picture of a pair of shoes, a purse or shirt in the world -- or in a store -- and find the very same or similar product online for sale. Amazon Remembers can do a basic version of this today," Sterling said. "I see lots of interesting applications for this across platforms."
Google Search Experiments
In other Google news, the search giant is testing a new feature that updates search results as users type search queries -- and modifies the results on the fly as the user continues typing. Some testers are complaining that it scrambles the page too often and can be confusing. It's not clear whether this experiment will become a permanent feature of Google's search experience.
Google also launched a change to its ranking algorithm that it said will make it easier for users to find a large number of results from a single site. The new feature debuted on Friday. So for queries that demonstrate a strong user interest in a particular domain, Google will show more results from the relevant site.
Google used "exhibitions at amnh" as an example. AMNH is the American Museum of Natural History. Before Friday's change, only two results from www.amnh.org would have appeared for this query. Now Google determines that the user is likely interested in the museum's web site, so seven results from the domain appear.
"We're always reassessing our ranking and user interface, making hundreds of changes each year," said Samarth Keshava, a software engineer at Google. "We expect today's improvement will help users find deeper results from a single site while still providing diversity on the results page."