Covering Disruptive Technology Powering Business in The Digital Age

Home > DTA news > Blog > Microsoft Invests in Deep Learning to Improve Image Search Results
Microsoft Invests in Deep Learning to Improve Image Search Results

 

Microsoft wants to make finding just the right image online less of a crapshoot by using deep learning to help its Bing Image Search engine return more relevant results to users.

Deep learning is a computationally intensive subset of machine learning that enables advanced analytical workloads. Using neural networks modeled after the human brain, deep learning systems can better mimic how people recognize patterns and process information.

Peeling back the curtain a bit on the search engine’s internal workings, the Bing Image Search Relevance Team at Microsoft revealed that they are using deep learning to unlock information held between the pixels of the billions of pictures found online.

Microsoft’s artificial intelligence (AI) searches for correlations between images and search queries mapped to semantic spaces used to derive meaning from data, even if the web pages containing those images lack any sort of text descriptors.

“This approach allows Bing to provide high quality results even from pages where no such query terms are present,” blogged Microsoft representatives. “We do this at internet scale, searching billions of images in tens of milliseconds for every image search query issued on Bing.”

Deep learning is also giving Bing’s non-image results a boost.

Microsoft recently disclosed details about the FPGA-based deep learning systems that power Bing’s new intelligent search capabilities, such as pulling information from various websites and presenting it to the user as a fact-filled summary. FPGAs (field programmable gate arrays) from chipmaker Intel allow Bing to analyze billions of documents on the web and deliver answers to the questions posed by users in a fraction of a second.

Meanwhile, Google is using AI not only to sharpen its own image search results, but also to predict which images humans will find aesthetically pleasing. In late 2017, the company researchers unveiled a new machine-learning model called Neural Image Assessment (NIMA) that can rank images based on how likely they are to catch a user’s eye.

More secure Bing Maps API Keys, improved intelligent answers

Developers using the Bing Maps API for geolocation capabilities in their own apps now have new security-enhancing feature. They can now restrict access to their API keys, which are code that is passed between applications to track and manage how an APIs is used, to create whitelists of domains.

If an IP address or referrer value falls outside the list, access is denied. For now, the feature’s referrer URL matching functionality is very exacting, but Microsoft revealed plans to support short URLs in the future. Also, IP address wildcards are not supported, but users can specify IP ranges that will work with the new feature.

Elsewhere on Bing, the Intelligent Answers feature is rolling out to more users in the U.S. and the U.K.  The feature will also display answers with up to five websites that back up the search engine’s findings.

And just in time for the busy summer travel season, hotel-related searches on Bing now gives users the option of opening open a new hotel-comparison view. It overlays room prices over a map, allows users to quickly compare properties and rates.

This article was originally published on www.eweek.com can be viewed in full

(0)(0)