The system can analyse up to 2,000 tweets a second, using complex software to extract from each tweet a direct expression of one of eight basic emotions*: anger, disgust, fear, happiness, sadness, surprise, shame and confusion.

As well as tracking initial public reactions to events, such as the 2011 riots and the murder of soldier Lee Rigby, the system can also analyse how the public mood changes over time following subsequent incidents or interventions.

There are many applications for the new system, from use by the police to track potential criminal behaviour or threats to public safety, to guiding national policy on the best way to react to major incidents.

Principal investigator Professor Tom Jackson from the University’s Centre for Information Management, based in the School of Business and Economics, says social media has enabled us to track very accurately how and what people are feeling.

“Twitter is a very concise platform through which users express publicly how they feel about a particular event, be that a criminal act, a new Government policy or even a change in the weather,” he explains.

“Through the computer program we have created we can collate these expressions of feelings in real time, map them geographically and track how they develop.”

Dr Ann O’Brien, who was part of the research team that created the ontology for emotions, adds: “The ontology we created takes the eight emotions and gives them a rich linguistic context so that we can chart the strength of emotions expressed in ordinary language and also in slang.  For any incident we can view how reactions grow and diminish over time.”

Researcher Dr Martin Sykora says “We have evaluated the system and have demonstrated its success of accurately capturing explicit emotions on an initial golden dataset.  The achieved F-measure, which is commonly used to evaluate performance of these types of systems, is currently the best that has been reported for this task.”

The system can be scaled up easily to monitor tweets globally, of which there are 10,000 a second. “Our program and the entire natural language parsing engine were optimised for big data, so that a fifth of all tweets coming through Twitter every second can potentially be processed on a single, average machine,” Dr Sykora adds.  “This copes with virtually all UK based Tweets, and adding further machines to the process can scale the analysis further.”

The research was funded by the Engineering and Physical Sciences Research Council and the Defence Science and Technology Laboratory.  For further information visit the EMOTIVE website.

*These are based and extended from Ekman’s basic emotions