People searching for news on Bing will now be given a complete overview of stories as part of a new feature launched in the UK today.
Spotlight will offer readers the latest headlines, a rundown of how the story has developed over time and relevant social media posts. Microsoft believes this gives Bing users a well-rounded view of a topic before they decide whether they want to learn more and click on an article.
“We can make sure that when we run a search, we find authoritative results,” said Jordi Ribas, a Corporate Vice President at Microsoft and Head of Bing. “For a question such as ‘is coffee good for you?’, there are some authoritative results that cluster around a positive sentiment, and some that cluster around a negative sentiment. Our algorithm can identify that and provide a multi-perspective answer.
“In my view, the role of a search engine should be to provide relevant, comprehensive and objective results. This will be a process of continual iteration and learning to make sure search results are as unbiased and as trustworthy as possible.”
Microsoft’s search engine documents signals from publishers, including how many publishers cover a story, their angles and how prominently they feature the story on their website.
For controversial topics, Bing shows different viewpoints from high-quality sources. For a source to be considered high quality, it must meet the Bing News PubHub Guidelines – a set of criteria that favours originality, readability, newsworthiness and transparency. Leading news providers identify sources and authors, give attribution and demonstrate sound journalistic practices such as accurate labelling of opinion and commentary.
Bing also uses deep-learning algorithms and web graphs of hundreds of millions of websites to identify top sources for news, per category, query or article.
“For web queries we are using machine reading comprehension, which means these algorithms are able to read the content of the internet, and find a paragraph that matches and answers that query,” Ribas added. “We are often able to promote that paragraph as an answer at the top of a page using deep-learning-based algorithms. We are able to surface these intelligent answers; sometimes it’s a single answer but sometimes the algorithm is able to detect that there are different points of view. The reason we are able to go deeper and give more objective and authoritative answers is because of this new generation of deep-learning algorithms.”
While the algorithms that make this all possible “go beyond anything we have done before in Bing”, Ribas said using those advances in technology to personalise the search engine to individuals could be a step too far.
“There is already an aspect of personalisation in Bing, because if you say ‘restaurants near me’ as a query, you want the search engine to know where you are. But we are cautious about it,” Ribas said. “Personalisation can go against objectiveness, because if you click on articles of a certain political preference, you will be getting more articles of that preference, because the algorithm is there to personalise based on your interests.
“We truly believe our mission is to provide results that are comprehensive, objective and as trustworthy as possible. We think personalisation is often not good for search engines because it could lead to biases.”
While Ribas has ruled out tailoring Bing to specific people, he said users will want to change how they use search engines so they can interact with them in a more natural way. That includes more searches using voice, which will turn into conversations.
“Ten years ago you would type in a query, and you would see a bunch of blue links,” he added. “Now, if you enter a query, you get a much richer experience and you often see a direct answer at the top. That’s because people’s expectations have increased over time. Rather than scroll through blue links, people expect to get a direct answer.
“As time goes on, we anticipate the user’s expectations to grow, not only in Bing but also from [AI assistant] Cortana. As time goes on, we expect the queries to be longer and more natural, while more of them will be completed using speech. Search will evolve from today’s query-answer model into more of a dialogue between the user and the search engine, and it will also evolve to become more visual as people search with images.”
Bing was launched in 2009 and now powers more than a third of PC web searches in the US and over 20% in the UK. It has evolved to offer accurate news reports, sports results, financial information, a dictionary and a translation service, on top of its core search function.
Earlier this year, Microsoft revealed that users could now search for information in Bing using pictures; so, taking a photo of a pair of jeans or a shirt in the Bing app will give you details about the item and where you can buy it.
All this is possible because of Microsoft’s extensive work in AI, which has seen the company emerge as one of the leaders in the technology.
