FACEBOOK STALKING MADE EASY

A university computer science professor and a graduate student have created the world's fastest computer program for finding patterns in social networks, which they say will be able to provide more useful searches on websites such as Facebook and YouTube.

The team's Cloud Oriented Subgraph Identification can connect 10 million to 20 million social network "nodes" — such as people's pictures, user IDs or profiles — in less than a second, according to computer science professor V.S. Subrahmanian, who led the yearlong project.

If implemented, COSI would let users search social networking sites using otherwise random search terms that would likely generate useless results with current search engine technology, said Matthias Bröecheler, the graduate student who assisted Subrahmanian.

For instance, Bröecheler said, COSI could help someone use Facebook to find a friend of a friend's cousin who enjoys kayaking and works at Burger King, or to find that video on YouTube everyone keeps talking about, without having to manually filter through unrelated profiles and other material.

"Currently, when we navigate social networks, all we can do is browse," Bröecheler said. "You can think of COSI as this little operator guy that zooms through the network trying to find the connections you specify."

In working on the project, the findings of which are set to be presented next month at a social networks analysis and mining conference in Denmark, researchers combined the program with existing "cloud computing" technology that uses multiple processors to answer complex multi-step questions. They hope to see COSI technology implemented on social networking sites within two years.

Although COSI isn't the first program to find patterns in social networks by linking nodes to "edges" — connections to nodes such as tagged photos — Subrahmanian said it's the first program that's able to connect so many. While he said existing programs can process no more than 15 million edges at one time, he said COSI can process up to a billion.

"The notion of a network is a very broad concept," Subrahmanian said. "When patterns get bigger and have more unknown edges, it gets increasingly harder to find what you're looking for."

Subrahmanian said if social networks are able to process huge amounts of data quickly with help from COSI, users may be able to find people or groups using search terms that current technology can't support. For example, on Facebook users would be able to find friends based on personal interests or hobbies using more generic keywords and still get specific results.

Subrahmanian said since the program is only beneficial when processing large amounts of data — at least eight search terms — social networking companies are far more likely than individual computer users to work directly with COSI.

But anyone who uses social networking sites will benefit if those sites choose to incorporate the technology into their search engines, he added.

"We think what COSI will do is provide companies with the ability to provide tools to make things easier for the average consumer," Subrahmanian said. "I'm sure social network companies could come up with a lot of ideas we haven't even thought of yet."

Senior computer science major Rayhan Hasan said he thinks the changes individuals notice will be helpful but subtle.

"There will be people at these social network companies who look at this program and find some cool app for it," Hasan said. "But we're used to being amazed by our computers, so people will probably just think their computer did something cool."

Senior biology major Elyse Geibel said the idea of making searches easier was intriguing, but she expressed concern about possible unintended consequences of the upgrade.

"I think the technology could be useful for social network users, but it could be just as useful for creepers," Geibel said. "I'm picturing some weird guy typing in ‘girl who likes Thirsty Turtle' and getting thousands of results."

Senior biology major Sarah Watt said the prospect of finding videos or friends with minimal effort outweighed the potential risks.

"Often, I really struggle finding a YouTube video if I don't have the right keywords," Watt said. "I think like with all new social networking things, it's something people will just need to get used to."

news at umdbk dot com