You have a big, shocking dataset.

When you have 1 billion users plus trillions of of their locations, objects, features, and social relationships, you'll find that you have a big, shocking dataset. So when the atlas was launched last month, Facebook stressed that they would use keyword search patterns.

The atlas search is defined by Facebook as an object-oriented search tool. On top of it, each person, every place, every item is a node with hundreds or even thousands of attributes that can be connected to each other.

Now, you can ask simple questions on the Atlas search to find friends who live in specific places or have specific hobbies. But so far the Atlas search has only been available for only 5 weeks, and its functionality is still very limited, according to Facebook, the full functionality of the Atlas search is now only 1%.

Therefore, the map search to achieve the next function is now the most concerned about. Of course, these features can be more complex and useful.

Let's get to know this from a computer science perspective. Edge markers, which are properties that show where you or other nodes will be connected, like your friends, people who live in Denver, people like Justin Bieber, and so on, they're just numbers, and each attribute gets a unique number (ranging from 1 bytes to 19 bytes). That's why we can just do a little bit of mapping and search for friends who live in Denver and like Justin Bieber. If you've read the Facebook Project blog published by Lars Rasmussen in how to Graph search works, you know that this is a central part of the technology that unicorn all Facebook searches.

This is a very concise search pattern. It will only show you the search results that have been established with you. It locks up search scopes based on common properties, such as what you like, where you work, who you know, and so on. It's like you're holding a metal detector in a lot of hay to find the needle you want.

"But what really pulls the wind is that we can help users find a bunch of new hay," said Facebook engineer Mike Curtiss at Facebook headquarters in Parahatu, Calif., "allowing users to search for things that haven't been connected to them yet." ”

What does Mike Curtiss want the Atlas search to accomplish? For example, if you are a senior student who is preparing to apply for a university, you can find out from the people who have gone to Harvard to give you some useful advice about applying for Harvard. If you are an HR, you can find out who has behaved badly among your job seekers through a map search. If you live in Denver and you want to taste the authentic Russian cuisine, the Atlas search will tell you about the Russian-style restaurants in Denver that were created in Moscow.

"Our search can ultimately answer almost all the questions, yes, all the questions!" Curtiss said. "But for now we can't do that because of all kinds of restrictions." But in theory, nothing can stop us from achieving that goal. ”

"The biggest problem at the moment is that the size of the answer set will rise exponentially," he said. If a node is connected to 100 other nodes, you will have 10,000 exit nodes. In the next round of search you will have 1 million exit nodes. This is a veritable problem, it is difficult to estimate, it is impossible to say that you add more than a few devices can solve it.

Facebook used the social ranking algorithm to determine which search results were most relevant or interesting before Facebook found a way to make hardware compress billions of nodes perfectly. This algorithm is a bit like the combination of unicorn, application operators, and EdgeRank. "It's like we provide a custom search engine for every user," Curtiss said.

When Facebook consolidates all of the open graph data together, the nodes contain far more than what you like to go to, including what you've recently bought, the pages you've commented on, and even your online game scores. The entire calculation process will be more complex, and the Atlas search will become smarter.

Now, the Atlas search is still a growing product and is only used by hundreds of thousands of of native English-speaking users. Compared to its final state, we can say that Facebook's Atlas search is only a baby.