A Deeper Look At Facebook’s Graph Search

By Justin Lafferty 

Not long after Facebook announced graph search, many people were mystified as to what the product was and how it could be used. Facebook invited a few journalists behind the curtain Thursday to discover just what makes graph search tick. The answer, naturally, is Unicorn.

As Facebook has more than 1 billion monthly active users and 240 billion photos, with 350 million photos, 2.7 billion likes, and 2.5 billion items of information shared each day, there has to be an efficient way to search through all that.

What makes graph search unique is a program called Unicorn. Facebook software engineer Mike Curtiss notes that it’s similar to a standard search engine, but different because it supports multihop queries issued in a series of steps. Within graph search, there are numbers that are connected to nodes — similar to keywords in a search engine. The nodes’ numerical structure connects them to the social graph.

Curtiss discussed a hypothetical where he’s moving to New York. Using Facebook’s graph search, he might want to find a new job in the Big Apple, and find friends who live there. Through the search engine, Curtiss could learn more about where his friends who live in New York are employed.

The way we would do that is within Unicorn. Start at the node for me, traverse the friend edge, and from those nodes, we can hop to the companies that my companies my friends work at through another edge.

CNET has a more in-depth look at how graph search works, if you’re interested in finding how you can search for friends who live in San Francisco and enjoy watching “The Big Bang Theory,” for instance. Facebook engineers noted that graph search is still a work in progress. As Soren Lassen, manager of the search infrastructure team at Facebook noted, the journey is just “1 percent” done.

Readers: What else do you want to know about graph search?