Microsoft unveiled a handful of intelligent search features at an event held in San Francisco on Wednesday. The news centered on powering services such as conversational and intelligent search with artificial intelligence, machine learning, and intent signals.
For example, Bing now validates its answers by sourcing several websites, not just one. It then validates more than one perspective in response to questions such as whether cholesterol is bad, so that both answers would serve up in the query results. When there is more than one way to answer a question, the answers will serve up in a carousel.
At the Microsoft event in San Francisco on Wednesday, executives showcased a number of advances in the company's Bing search engine, Cortana intelligent assistant, and Microsoft Office 365 tools. All use artificial intelligence (A.I.) to help people get information and assist with complex requests.
Steve Sirich, general manager on the global search ads business at Microsoft, describes one of these complex requests as the ability for the search engine to put numbers into perspective. For example, rather than return a query telling the person searching that Syria is 71,498 square miles, the results will compare it to the size of the U.S. state of Florida.
For people who are watching their caloric intake, when someone queries "how many calories are in a serving of ice cream," they will learn that a scoop contains 137 calories, which is equal to about 11 minutes of running.
Bing's expertise in web-scale mining of billions of documents and with conversational A.I. with chatbots like Xiaoice and Zo has created a new way to search that is interactive and can build on previous searches to get people the best answer.
Now, when someone searches for specific topics related to health, Bing will ask clarifying questions to better refine search when needed.
Sirich said Bing and Bing Ads are the largest A.I. applications at Microsoft. "A.I. is really fueled by data," he said. "When we set out in 2009 to launch Bing, we were continuing to drive into the A.I. frontier because of our ability to index and create a digital canvas."
When Bing engineers think about A.I. and Bing Ads, three areas come to mind, such as audience. Search is moving from the use of targeted keywords to audience buying. A.I. powers Bing's audience graph. The data comes from the 600 million devices running Windows 10, Office 365, and LinkedIn.
Microsoft recently made Dynamic Search Ads on Bing generally available to all advertisers, Sirich said. For example, about 15% of queries on Bing are new monthly. "That represents more than one billion queries a month that we and advertisers have never seen before," he said, explaining that advertisers opting into DSA now enables the message to match more precisely with the query.
Other new features in Bing that use A.I. include multi-site support, where Bing highlights the query when an answer has been seen across the web. It combines recent breakthroughs in machine reading with the Bing knowledge graph to identify how many reputable sources an answer has, giving users more confidence that the answer is correct. This also saves time that would have been used in clicking through to validate the answer themselves.
Bing also now clearly compares two entities based on relevant characteristics such as when people want to know the key differences between two similar items like a crow and a raven.
Microsoft also shared details on Bing’s advanced image search features. Bing Image Search leverages computer vision and object recognition to provide more ways to find what people are looking for. Users can search any image or within images to shop for fashion or home furniture. Bing detects and highlights different products within images, and the user can click the magnifying glass icon on the top right of any image to search within an image and find related images or products.