A new system for creating trusted recommendations online has attracted interest from Internet giant Yahoo and earned its SFU developers a best-paper award at an international conference in Spain last month.

The paper, by computing science doctoral candidate, Mohsen Jamali and computing science professor Martin Ester, won out over some 200 other international submissions at the ACM (Association for Computing Machinery) Conference on Recommender Systems.

Recommender systems, such as those used by video-rental site Netflix and online merchandiser Amazon, help users find relevant information on products or services based on how they have rated previously acquired items.

The systems are becoming increasingly important for navigating through the flood of information on the Internet.

But Ester says existing systems are not useful for those “cold-start” users who are new to the system and have rated only a few items. “This is a serious problem, since in real-life systems as many as 50 per cent of users tend to be cold-start users.

Jamali and Ester’s alternative system exploits social networking sites because it assumes that users will trust their friends’ ratings.

“We explore direct and indirect neighbours of the target user in a social network and collect their highly rated items to make a recommendation,” explains Ester.
“This approach also works for cold-start users, as long as they are somehow connected to a social network.”