Dataminr for News aims to help reporters spot breaking stories in a sea of tweets

作者Maite Fernandez
Feb 3, 2014 发表在 Miscellaneous

As Twitter, the preferred social network for breaking news, keeps growing, it can be a challenge to find the signal in the noise.

In the stream of the latest tweets, the latest cat gif is followed by an Instagram sunset, a photo from the Ukraine protests and Justin Bieber’s mugshot. Being able to find the quality information about breaking news amid 500 million daily tweets gives a reporter an important edge.

Enter Dataminr for News, a new tool developed as a partnership between Twitter, CNN and social media analytics company Dataminr. Dataminr technology solves Twitter's information overload problem "by algorithmically discovering, qualifying, categorizing and alerting clients to key information in real-time,” explained the company in a press release.

The New York-based data analysis company has been doing this for the financial sector and government agencies since 2009. Now Dataminr has partnered with Twitter and CNN, which has been testing the tool for a few months, to push it out to newsrooms and journalists.

But how does Dataminr unearth key information from the noise? In a presentation at DevNest, a Twitter developer meetup, CEO Ted Bailey explained that Dataminr uses about 18 different signals derived from Twitter and its metadata to help identify newsworthy information and to create customized alerts.

One emblematic case was the 2011 death of Osama bin Laden, one of the most-tweeted events in Twitter history. Dataminr sent an alert at 10:20 p.m. Eastern time, 23 minutes before it was reported by news outlets. In that case, Bailey explained, they used three signals (message volume, sentiment classification and linguistic analysis) based on 19 tweets posted before the news was confirmed.

The volume of tweets alerted Dataminr that a hot topic was emerging. While 19 might seem like a low number, bin Laden hadn't been a hot topic recently, explained Bailey, so that number of tweets at that moment was abnormal.

That sudden interest in bin Laden caught their attention, but didn't yet justify sending an alert, so they looked at linguistic analysis. They also looked for--and found--sentiment, which Bailey defined as the presence of a sudden “emotional reaction."

Once newsworthy information is detected, it’s sent to users via the app, email alerts or text messages. The alerts can also be built around their editorial workflow.

Dataminr for News, which will be a paid service, will be released this year.

Photo: CC-Licensed, thanks to Garrett Heath on Flickr.

Maite Fernández is IJNet’s managing editor. She is bilingual in English and Spanish and has an M.J. in multimedia journalism from the University of Maryland.