A study conducted by researchers from Cardiff University in the United Kingdom has suggested that data from social media platforms such as Twitter is capable of identifying dangerous situations faster than reports by the police. The difference in detection speeds was found to be in some cases one hour faster.
The researchers used 1.6 million tweets as a data set. The sample tweets were sent out during the London riots that occurred in 2011. Using machine learning algorithms, the tweets were scanned with a view to identifying the threats. Factors such as the tweet’s location, the timing of the tweet and how frequently tweets containing particular words or variations of these words were also considered.
Machine learning algorithms
When the machine learning algorithms were applied to the case of the London riots, the researchers were able to tell incidents earlier than the police in just about every instance. In one example, the summarization by the system of a tweet reading, “not feeling the rumors that the rioters are looking to move to edmonton and #enfield town. DON’T YOU PEOPLE THINK YOU’VE DONE ENOUGH!!!!” was detected almost half an hour earlier than the police could receive intelligence that certain members of a known gang were planning to go to Edmonton in order to cause chaos.
The report also notes that current approaches of detecting dangerous events or incidents are heavily invested towards big events such as terror attacks. It is thus harder to get information on smaller incidents such as car accidents or fires. But by leveraging data obtained from social media, this resource gap can be closed.
Augmenting police intelligence-gathering
Data from social media can also be used in large-scale incidences to enhance the efficiency of the response or to manage the dispensation of the right information. With social media data it is not only possible for the intelligence-gathering techniques of the police to be augmented but it makes it easier for the police and the general public to pick up on smaller incidents to predict dangerous situations such as riots prior to their occurrence.
The Cardiff University researchers who participated in the study were Omer Rana, Pete Burnap and Nasser Alsaedi. Their research report was published last month.