News & Analysis
Comment
utilitus
This article dramatizes how little real engineering goes on at Facebook. ...
IDontUseTheForumSoWhyAmIForcedToMakeANickname
I traveled extensively around the world and the first rule is: Never visit a ...
Facebook's ex-Google guy talks social search
Rick Merritt
1/16/2013 12:30 AM EST
How Graph Search works
Google already had a cross-indexing feature under the covers it used for basic inquiries such as delivering news about which of your friends have birthdays today. Rasmussen and his team saw the potential to radically expand such cross-index searches.
At a high level, the software has two components, both built in house over the past year. One is a natural language querying program.
Rasmussen and his team evaluated several existing natural language tools. They decided to build their own to give themselves flexibility.
Underneath the natural language interface lies Unicorn, an open source program to cross-reference indexes. The Facebook team tries to use mainly open source tools such as Unicorn and HBase. The new Graph Search feature does not use the open source No SQL database popular in many Web 2.0 companies, he said.
Rasmussen does his work at relatively high level of code. He readily admits he does not know how Graph Search interfaces with lower levels of Facebook’s software or what if any implications it has for the balance of storage and compute in its data centers.
The Graph Search team included “tens of people,” Rasmussen said. Most teams at Facebook are much smaller, involving just a handful or people focused on moving fast. Indeed another slogan pasted on the walls here is “Move fast and break things.”
Will Graph Search be another Google Wave, a belly flop in the Web 2.0 evolution?
Rasmussen said he created Wave for himself, as a reflection of his needs as a heavy email user at the time. In hindsight, he says, he didn’t validate his needs with the broader market.
That won’t be the problem for Graph Search, he said. A Facebook lab team has already been hammering on the code and will be listening to the first round of beta testers carefully
“I don’t recommend failing, but if this fails it will be for another big reason,” Rasmussen said.
Zuckerberg called Graph Search a third pillar for Facebook.
Next: Facebook vs. Google
Google already had a cross-indexing feature under the covers it used for basic inquiries such as delivering news about which of your friends have birthdays today. Rasmussen and his team saw the potential to radically expand such cross-index searches.
At a high level, the software has two components, both built in house over the past year. One is a natural language querying program.
Rasmussen and his team evaluated several existing natural language tools. They decided to build their own to give themselves flexibility.
Underneath the natural language interface lies Unicorn, an open source program to cross-reference indexes. The Facebook team tries to use mainly open source tools such as Unicorn and HBase. The new Graph Search feature does not use the open source No SQL database popular in many Web 2.0 companies, he said.
Rasmussen does his work at relatively high level of code. He readily admits he does not know how Graph Search interfaces with lower levels of Facebook’s software or what if any implications it has for the balance of storage and compute in its data centers.
The Graph Search team included “tens of people,” Rasmussen said. Most teams at Facebook are much smaller, involving just a handful or people focused on moving fast. Indeed another slogan pasted on the walls here is “Move fast and break things.”
Will Graph Search be another Google Wave, a belly flop in the Web 2.0 evolution?
Rasmussen said he created Wave for himself, as a reflection of his needs as a heavy email user at the time. In hindsight, he says, he didn’t validate his needs with the broader market.
That won’t be the problem for Graph Search, he said. A Facebook lab team has already been hammering on the code and will be listening to the first round of beta testers carefully
“I don’t recommend failing, but if this fails it will be for another big reason,” Rasmussen said.
Zuckerberg called Graph Search a third pillar for Facebook.
Next: Facebook vs. Google
|
|
|
|
|
Navigate to related information
iniewski
1/16/2013 12:21 PM EST
I don't think this will work...it will be another flop as Waive
Sign in to Reply
Robotics Developer
1/16/2013 2:58 PM EST
Iniewski, I agree with you! I have been moving away from Facebook use more and more each day. I can't imagine needing to graphically search my friends opinions rather than ASK them (or already know!). This seems to be a non-starter search engine aimed possibly at those who don't interact with their "friends" or don't want to. Maybe this will be next big "Waive" but I am not holding my breath.
Sign in to Reply
elPresidente
1/17/2013 2:50 AM EST
Google's search engines have totally sucked as a useful tool over the past couple of years.
You can have your "personal" road for search - as an engineer, I don't want anything do with anything that's not UNBIASED, unpaid-for, facts and, preferably, is as socially devoid of BS as the machines I build.
Sign in to Reply
rick.merritt
1/17/2013 2:53 AM EST
I see two sides of this coin:
Facebook has some wonderfully personal data sets tied to the hearts and minds of its users.
But those data sets are inherently subjective, limited to Fbook users.
In any case, I am interested in the software workloads that are driving 1) the cloud and 2) software engineering and Facebook is one of a handful of the big drivers.
Sign in to Reply
rick.merritt
1/17/2013 3:15 AM EST
Search Graph is shrewd in the way it encourages more users to share and put meta data on more of their stuff, in the process increasing the use and value of Facebook's database.
Sign in to Reply
IDontUseTheForumSoWhyAmIForcedToMakeANickname
1/17/2013 12:12 PM EST
Well put, I see it more of a way to monetize personal searches. If I can strategically place an ad/link for a product that one of my friends has talked or posted about, then chances are I'm more likely to click on it. And since facebook really has no real revenue source, you can bet your shareholder equity that this is a no brainer.
Facebook has the unique position of OWNING all the personal data that you're willing to post about yourself. Age, maritial status, sex, location, likes, dislikes, friends, hobbies ... you name it. It's a marketer's wet dream. Monetizing that information is one of the biggest reasons for facebook having its IPO. Social Graph is merely one step in the process.
Sign in to Reply
Rick@DassaultSystemes
1/17/2013 12:54 PM EST
Thinking beyond the initial release of this function, it looks like Facebook is venturing beyond the search into "Search-based Applications". I've been learning about this through Exalead, a company that was acquired by Dassault Systemes (where I work). It's been very interesting learning about SBAs. What I've learned from experience and from customers over the years is that getting data into a system is one problem, but the key value is how to get information out of a system. The more intelligence that can be added to that process, the more useful the data. Pretty cool and useful stuff. Feel free to check in with the Exalead team. But I applaud the efforts.
Sign in to Reply
selinz
1/17/2013 1:05 PM EST
Sounds very interesting.... The interesting aspect is how much user data will be "spread."
Sign in to Reply
Frank Eory
1/17/2013 4:00 PM EST
There is great potential here, but it sounds like it may take years to realize that potential.
Graph Search does however, seem like an essential "third pillar" for Facebook. One of the biggest problems with the timeline & newsfeed is their linear chronological nature. In theory, one could scroll down indefinitely through his newsfeed or a friend's timeline to look for a past event of interest, but that is far too cumbersome and time-consuming for most people.
In addition to the examples mentioned in the demo, vacation travel is another good one that comes to mind. If I'm planning a trip somewhere, I would find it very useful to be able to search my FB friends to find out who has been there, where did they stay, etc., and then contact them for advice and recommendations. Such personal recommendations from trusted sources are an advertiser's dream.
Sign in to Reply
przem
1/19/2013 8:43 PM EST
@Frank---absolutely right. I always asked what do the locals do, and followed there.. I ask my friends and recently scour places like Reddit or Facebook what _they_ like to do in their town because, obviously, they aren't going to follow standard guide advice.
Sign in to Reply
IDontUseTheForumSoWhyAmIForcedToMakeANickname
1/21/2013 10:04 AM EST
I traveled extensively around the world and the first rule is: Never visit a restaurant that's advertised in the booklet left in your hotel room (unless you want a so-so meal).
I'll ask at the hotel's front desk where's a good place to eat. When they come back with the standard recommendations from the hotel book, I immediately follow up with 'Where would you personally eat?' and the responses are ALWAYS different. Locals tend to know the best meals, deals & places to go. Brochures only promote those that paid to attract business. Great places don't need to advertise because word of mouth will keep them open or close them down. Locals are the key because they live there and visit these places themselves.
Sure, I could check someones facebook page and hope they recommended something in the town I'm visiting ... call me old fashioned, but I'd rather talk to them directly about the full experience. That way, I can better understand why they rated it high or low.
Sign in to Reply
nannasin28
1/21/2013 3:29 AM EST
then chances are I'm more likely to click on it. http://www.hqew.net
Sign in to Reply
utilitus
1/22/2013 2:40 PM EST
This article dramatizes how little real engineering goes on at Facebook. Zuckerburgs' Erdös number is infinity.
Sign in to Reply