Communications of the ACM
Keeping Pace With the Data Explosion
Two researchers have developed co-factorization machines in order to study the interaction between social media users and tweets.
Credit: Lehigh University
Two Lehigh University researchers have developed co-factorization machines that utilize mathematical analysis to study the interaction between social media users and tweets. "If we can better understand what you are interested in, we can decide what to filter, rank higher or flag for your attention," says professor Brian Davison with Lehigh's Web Understanding, Modeling, and Evaluation lab.
Davison and fellow researcher Liangjie Hong at Yahoo Labs analyzed a surge of Twitter activity and then trained an algorithm to predict with high accuracy how frequently the tweet recipients would retweet the messages to their own followers. The method also reveals users' possible interests according to factors that include the regularity with which specific terms appear in their feeds. Davison and Hong's algorithms improve their predictions via machine learning, as the programs learn from the outcomes of past interactions and form a basis of rules for individual users.
Davison expects algorithms will soon tailor a social media user's experience, while Hong notes users' response to information and their decision to pass it along or not is typically shaped by a combination of personalization and popularity. "We need to offer you the information that is most relevant to you while providing the popular stuff," Hong says.
From Lehigh University
View Full Article
Abstracts Copyright © 2014 Information Inc., Bethesda, Maryland, USA
No entries found