While we've seen work on an algorithm that can predict trending Twitter topics before they happen, that's no good if you're sifting through a mountain of tweets to find accurate, up-to-the-minute news. For that, you might want to look into a new paper to be published in Internet Research next month and reported on by Slate. In it, researchers test out an algorithm designed to assess the accuracy of information posted on Twitter.

The machine-learning system uses a number of parameters based on prior research which found typical signifiers that a tweet might contain valid information — for example, if it comes from an account with a lot of followers, if it contains a link, or if it's negative in tone, it's more likely to be true.

Right now the algorithm apparently has an "area under the curve" (AUC) of 0.86, meaning it's closer to a perfect score of 1 than a random score of 0.5. Of course, there's a lot of room for improvement, but the potential is clear; with software able to work a lot faster than humans, such a tool could help the more valuable information on Twitter rise to the top in emergency situations.