Friends and Experts


The foundation of web search has been built on keywords, links and clicks pointing to pages. This approach is great for finding web sites but search is more than about simply finding pages. With the help of social networks, people are able to share nearly everything they do in digital form and offer their opinions on almost every conceivable topic. From real-time streams to social conversations, connections are created that present the opportunity to bring people into the search equation. Today Paul Yiu, Principal Group Program Manager for Bing Social, provides an overview of how we incorporated people into our latest release. No matter what query you submit to Bing, you may be amazed that some of your friends, influentials or experts often know something that you are searching for, in addition to high quality web documents that you always count on.

– Dr. Harry Shum, Corporate Vice President, Bing R&D

people sqi

The goal here is to bring you a list of people that might be able to help you get more done. For instance, here is an example of how the Sidebar might help you plan the perfect adventure in Hawaii.

Friends Who Might Know

Depending on what your friends have been paying attention to, and their profile information, Bing will recommend them as people that might be able to help. Let’s look at the Hawaii example. For me, when I search for “Hawaii,” I see the following:

– Two friends that have shared beautiful photos of Hawaii

– A friend who likes the Hawaii page on Facebook

– A friend who likes a couple of topics related to natural aspects of Hawaii, and

– A a friend who lives in Honolulu, who could help me plan a trip

Another example of how this works is with general category search. For example if you search for movies or restaurants, you may see things your friends have liked, as potential recommendations. I was delighted to find out how much I have in common with friends in terms of our taste in movies and food.

In terms of how we order these friends in the sidebar, it’s a combination of how many activities and attributes match your query, the type of activities and attributes that made the friend relevant, and how likely our ranking system thinks you will find that information useful. Since we launched the feature a month ago, the system is learning quickly which types of information inspire the most engagement from users. The more you use the product, the more accurate that Bing gets at recommending friends that might be able to help.

People Who Know

Beyond friends, Bing can help you find people who are influential about the topic you’re searching, based on what they’ve publically blogged or tweeted about. In a glance, you will see top experts and enthusiasts from leading networks like Twitter to quickly check out what they have to say about the topic you’re searching for. You can follow them, ask them a question or see what they have shared in the past. While results may vary when it comes to Friends Who Might Know, in the People Who Know section of the Sidebar, for now, Bing displays the same results for each user. The idea is to recommend people that are influential or popular in the context of your query or topic.

There is some similarity here to how we think about ranking documents. There are signals that are relatively static, and there are signals that are more dependent on the query. In terms of static signals, we look at:

– Followers in Twitter, and how many there are

– How influential the person is in general, i.e., how much does he or she get re-tweeted

– Who he or she follows on Twitter

– The likelihood that the Twitter user is a spammer based on peculiarities in his or her connectivity graph.

When it comes to query-dependent signals, we look at a user’s influence, i.e., how well does his or her content get retweeted around this particular topic. In a way, a retweet is like “social anchor text.”

Here is an example. I was interested in finding out more about a new HBO show by Aaron Sorkin, so I searched for “The Newsroom.” I see that Emily Nussbaum, a TV critic from the New Yorker, has been engaged in interesting conversations on Twitter. The content on Twitter supplements really well the news articles and reviews I can see in the web search results.

While general authority matters, our machine learning techniques try to surface the people who are Influential or Popular for the particular query topic at hand. As a result, people who are generally influential on Twitter (i.e. have a large general following) may not be guaranteed to appear for a query or topic where he or she has little influence. For example, Kim Kardashian is influential on Twitter but probably won’t appear for the query “machine learning.”

More to Come

We are working feverishly to enhance the content and improve the quality of people results. We’ll let you know as soon as new capabilities are available. Many thanks for your helpful comments and suggestions. Please keep them coming!

– Paul Yiu and the Bing Social Search Team