Algorithm Predicts Which News Tweets Will Be Popular: STUDY

How To Predict Which Tweets Will Go Viral

A study from UCLA and Hewlett-Packard's HP Labs suggests that it's possible to predict the likability of a news tweet before it's even released into the digital universe. Researchers used a prediction algorithm that was 84 percent accurate when mapping out the future of a post on Twitter.

Using the study as a model, The Atlantic pulled the following tweet from The New York Times' Twitter account, suggesting that this post is bound for popularity:

But what makes this somewhat unexciting tweet so shareable?

According to the study's release on the HP corporate Data Central blog, authors Roja Bandari, Sitram Asur and Bernardo Huberman developed their theory by considering four factors of a news story that might impact the success of a tweet linking to an article.

First, the researchers were concerned with which news sources created and published a story, and if credibility factored into the story's appeal. Second, they looked into the popularity of different categories of news, such as health, money or sports. Next, researchers asked whether the article's tone was emotional or objective, and would the audience veer toward one or the other? Finally, the news articles were searched for mentions of "celebrities, famous brands or other notable institutions."

The research paper explains how data was gathered from over 40,000 news articles throughout a nine-day period in August of 2011. Using scoring functions from within publicly available news feed aggregators, such as Feedzilla, the researchers were able to score articles based on the above four factors.

Bandari, Asur, and Huberman found that some news topics fell into much more popular categories than others. "Technology was the most tweetable news area followed by Health and the ever-shareable Fun Stuff," The Atlantic reads. Technology news was also often shared by popular websites such as Mashable and the Google Blog, which helped add to this category's popularity.

The Stanford Named Entity Recognizer was also used to "identify test representing a person or company name, and the historical prominence of the person or company on Twitter," according to HP's blog. In basic terms, that means if Justin Bieber's date is discussed in a tweet, the likelihood of the tweet's success is much greater.

Contrary to editorial belief, the tone of an article had little to do with its popularity. An objective or emotional voice within a story didn't matter nearly as much as the topic of conversation -- or as much as the creator of the content.

"Overall, we discovered that one of the most important predictors of popularity was the source of the article," the research paper reads. "This is in agreement with the intuition that readers are likely to be influenced by the news source that disseminates the article."

In some ways this makes sense: Celebrity stories from credible news sources tend to get passed around a lot. Yet interestingly enough, the tone of the news and the tweet linking to it (no matter how many splashy adjectives are used) has little to do with popularity. This study is also fascinating in that their predicted algorithm has such a high success rate and could be extremely applicable to news syndicates and avid Tweeters -- which, during an election year, could make a big difference.

"…Activists and politicians are increasingly using social media to influence public opinion. By testing their messages using our algorithm, they may be able to improve the visibility of their cause,” says Bernardo Huberman, an HP Senior Fellow and co-author.

Why do you click a news story? Do you believe the above four factors influence your decision, and if so, which is most important? Let us know in the comments section or tweet us (because we want to be popular too) at @HuffPostTech.

For a quick overview of this Twitter report from UCLA and HP Labs, check out
Go To Homepage

Before You Go

Sheryl Sandberg

50 People In Tech To Follow On Facebook

Popular in the Community