Facebook’s graph search has been hailed as a brand new feature, one of what CEO Mark Zuckerberg called a “pillar” of the social network along with its personalized News Feed and timeline profiles.
In fact, the search function is more or less a revival of an older feature that Facebook killed off, according to former Facebook executive Dave Morin.
Morin, now CEO of the mobile social network Path, says TechCrunch’s Mike Butcher at the DLD conference in Germany that Facebook had an “Advanced Search” function during the time he worked there, from 2006 to 2010.
According to Morin, this search function allows you to search for “my friends from SF who [sic] ski” – a query very similar to natural language searches that Graph Search beta users can now perform. (Facebook has limited adoption of Graph Search.)
Here is a screenshot from Kagan’s blog. (This version doesn’t seem to have the same free-form language research that Morin mentioned.)
The problem, Morin suggested, was scalability — the search function’s ability to keep up with computational demands as Facebook’s user base exploded into the hundreds of millions in the latter part of the past decade.
Facebook CTO Lars Rasmussen’s explanation of how he built Graph Search also supports Morin’s observation. Facebook had maintained three separate systems to look up its user data, Rasmussen said. One of these systems, Unicorn, has been upgraded to handle all kinds of searches and now forms the main basis of Graph Search.
There are traces of these old search systems on the site today. For example, an old web address for Facebook search is facebook.com/srch.php. This redirects to a search page focused on people who lets you search for friends based on location, employer, or school.
Another old web address, facebook.com/search.php, leads to current search interfacewhich sorts the results by people, Facebook pages, places, events and other data on the social network.
The big innovation in Graph Search is the new user interface on top of that data, which takes plain English sentences and translates them into structured queries that Facebook’s back-end search systems can handle. This is a big improvement over the state of Facebook search in recent years.
But Graph Search only tracks limited types of “structured” data – the kind of information you enter into your profile fields, report by clicking a like button, or attach to photos or status updates, such as friends’ names and locations. Graph Search does not perform free-form searches on words and phrases in status updates, for example. In that way, it’s very similar to the old system that Facebook killed off the last decade.
In an intriguing coincidence, considering how closely senior Facebook executives monitor Morin and his company, Path launched its own search feature last month.
A Facebook spokesperson confirmed that Morin’s account of the search feature’s history was largely correct.