What makes a tweet go viral online? And what does a viral trend actually look like?
Those are a couple of the questions that can be answered by a new Microsoft Research project, called Viral Search. The company’s researchers are showing the project this week at an internal gathering this week in Redmond. The program crunches large amounts of data from Twitter (and potentially Facebook and other platforms in the future) to analyze and display patterns of distribution on the social network.
See a video demo of the project above. Jake Hofman, a researcher from Microsoft’s New York lab, explained that popularity doesn’t necessarily mean that a piece of content is viral. For example, a large media organization can make a piece of content popular simply by essentially broadcasting it to millions of Twitter followers. But true virality means that a wide variety of people are tweeting independently and spreading the content in parallel.
Microsoft’s program shows what those trends look like. It can also dive in to show the impact of specific people and tweets.
This type of analysis is difficult given the sheer amount of data, Hofman said. “We have billions of relationships between individuals, and every day hundreds of billions of tweets,” he said. “Actually doing this at scale and sorting through these things requires a reasonable amount of computational power.”
Microsoft is talking about ways to possibly incorporate this into products, but for now it’s purely a research project.