Skip to main content



News Waves on Facebook

Facebook algorithms used to miss or react slowly to news waves or topics of current interest that were spreading quickly. For example, if a passenger airplane operated by a major international airline disappeared overnight, you may not see related content posted by organisations and friends towards the top of your Facebook news feed. To address this shortcoming, Facebook is employing attributes such as “when people are choosing to like, comment and share” and “the rate of interactions, and when they ramp up or taper off” to promote news wave topics towards the top of the news feed. Local attributes such as liking, commenting and sharing posts can be assessed using two Facebook friends while global attributes, including the rate of interactions and the change in the rate over time, need to be evaluated across many Facebook users.

If we create a network or graph with nodes as people on Facebook and edges between two nodes representing a friendship relation, we can analyse recent interactive activity between two Facebook friends with respect to a particular topic by assigning a topic interactivity score to each edge. Assume that a web crawler extracts recent news articles and an unsupervised topic model, such as probabilistic latent semantic analysis, is used to identify keywords in those articles that correspond to a particular topic. The topic interactivity score can then be computed by counting the number of liked, commented or shared posts between those two friends with keywords relevant to the topic. The global attributes can be evaluated by looking at how posts with keywords relevant to the topic spread through the network and the rate of transmission.

How can Facebook’s algorithms be improved to prioritise news wave posts in news feeds? Perhaps we can combine local information on the topic interactivity and global information on the structure of the network. Granovetter’s paper on the strength of weak ties highlights that new information comes from local bridges in networks where a local bridge is defined as an edge whose endpoints have no neighbours in common. If news feed ranking algorithms weigh topic interactivity information from local bridge edges more heavily, there should be more news wave content towards the top of news feeds.

Sources:

Main articles

Relevant papers

 

Comments

Leave a Reply

Blogging Calendar

September 2014
M T W T F S S
« Aug   Oct »
1234567
891011121314
15161718192021
22232425262728
2930  

Archives