Tracking public anger via Twitter

Nieuws | de redactie
17 april 2012 | A team of researchers from Bristol University tracked the British public mood by analyzing Twitter messages between 2009 and 2012. Budget cuts and the period leading up to last year’s riots triggered particular unrest. The Royal Wedding smoothed it over.

New research has analysed the moodof Twitter users in the UK and detected various changes in themood of the public. In particular, the researchers observed asignificant increase in negative mood, anger and fear, coincidingwith the announcement of spending cuts and last summer’s riotstogether with a possibly calming effect during the royalwedding.

Mood tracking via Twitter

In this study the researchers focused on measuring the mood, andchanges, using standard tools for mood detection, of a large sampleof the UK population. A collection of 484 million tweets generatedby more than 9.8 million users from the UK were analysed betweenJuly 2009 and January 2012, a period marked by economic downturnand some social tensions.

The findings present intriguing patterns that can be explainedwhen events and social changes are taken into account. Theresearchers found that a significant increase in negative moodindicators coincided with the announcement of the cuts to publicspending by the government, and that this effect is still lasting.They also detected events such as the riots of summer 2011, as wellas a possible calming effect in the run up to the royal wedding.Intriguingly, a rise in public anger seems to have already beenunder way in the days before the riots.

Large data set

Nello Cristianini, Professor of Artificial Intelligence,speaking about the research, said: “Social media allows for theeasy gathering of large amounts of data generated by the publicwhile communicating with each other.

“While we leave the interpretation of our findings to social andpolitical scientists, we observed how the period preceding theroyal wedding seems to be marked by a lowered incidence of angerand fear, which starts rising soon after that. Of course, otherevents also happened in early May 2011, so they may also beresponsible for that increase.”

The aim of the study was to see if the effects of social eventscould be seen in the contents of Twitter. The first part of theresearchers analysis provides a sanity check, in that itcorroborates their assumption that word-counting methods canprovide a reasonable approach to sentiment or mood analysis.

Happy days, angry days

While this approach is standard in many applications, theresearchers felt that a sanity check in the domain of mooddetection via Twitter was necessary. By making use of lists ofwords that are correlated with the sentiments of joy, fear, angerand sadness, they observed that periodic events such as Christmas,Valentine’s Day and Halloween evoke the same response in thepopulation, year after year.

The main part of the analysis focused on a visible change-pointoccurring in October 2010, when the government announced cuts topublic spending, testing its statistical significance. The studyshows that the change point is real, and that its effects can stillbe observed. In other words, public mood has still not recoveredfrom that announcement.

Forecasting the riots?

The same testing technique shows another important period, thatof summer 2011, when riots broke out in various UK cities, leadingto looting and even loss of life. The researchers method seems tosuggest that some increase in public anger preceded, and notfollowed, these events, although the significance of this findingis a matter for social scientists to discuss.

Future work will include the comparison with social mediacontent with traditional media content, as well as the comparisonof both with traditional opinion polls methods.

It is important to remark that the real-time detection of socialtrends via the analysis of social media content, presents variouspossible limitations.  Social media analysis can only beaccomplished with text mining technologies, which are less accuratethan human assessment, but can be applied to vast amounts of data.Also the population that is assessed is necessarily that of Twitterusers, which is a biased subsample of the general population.Particular care needs to be paid when extracting information butalso when reporting it.

Schrijf je in voor onze nieuwsbrief
ScienceGuide is bij wet verplicht je toestemming te vragen voor het gebruik van cookies.
Lees hier over ons cookiebeleid en klik op OK om akkoord te gaan