Twitter has repeatedly proved its usefulness as a platform for sharing information during times of social crisis. Now researchers from the University of Cardiff have demonstrated the ability of the service to predict serious incidents significantly faster than the police. In some cases, this difference reaches one hour.

Researchers analyzed data on riots in London in 2011 and showed that the computer can automatically scan Twitter and identify potentially dangerous events. The system, for example, would be able to report broken windows and set fire to cars even before the London police knew about it.

The computer can analyze rumors on Twitter and determine where the likelihood of riots is greatest. The system also knows how to report in real time large clusters of people.

Researchers used a number of machine learning algorithms to analyze 1.6 million tweets. The system took into account the time and place of publication of messages, as well as their contents. The algorithms found the incidents faster than the police in all cases, except for two.

One of the authors of the study, Dr. Pete Burnap (Pete Burnap), believes that Twitter monitoring technologies can help in the fight against future unrest.

"Previously, we processed Twitter data using machine learning and natural language systems to better understand online deviations such as spreading hostile statements and cyber-hatred," he said..

"In this study, we show that social networks are the starting point for reporting everyday events, including social unrest and ground-based criminal activities," Burnap continued.. "We will never replace traditional police ground resources, but we have demonstrated that this study can improve existing data collection systems and allow the creation of new technologies to support established policing practices".