Thursday, May 5, 2011
Visualizing opinions, emerging trends and finanical markets
On 23 January 2011, Portugal held its Presidential elections, re-electing Aníbal Cavaco Silva to a second term. Cavaco Silva won against all other candidates, getting 53% of the vote. But that’s not all there is to the story. There was a surprise hidden in amongst the candidates. Fernado Nobre, a Portuguese doctor who has participated in almost 200 humanitarian missions, was running as an independent. Without the support of any of Portugal’s major political parties, no one suspected he would be popular. So when Nobre collected 14.1% of the vote – more than the traditional candidates supported by the Communist Party, the New Democratic Party and the Greens – everyone was surprised.
Everyone, that is, except Jorge Louçã. Louçã is a computer scientist at Lisbon University Institute who studies ‘opinion dynamics’ on the Internet, by analyzing the content of newspapers, scientific journals, prominent blogs and social media. For two months leading up to the election, Louçã was studying all mentions of the presidential candidates. He knew that while none of the traditional media was covering Nobre, his name was going wild on Facebook. He was clearly popular despite a lack of media coverage.
Louçã is now part of the multi-institutional FuturICT (pronounced “futurist”) project, which aims to understand and manage complex, global, socially interactive systems. His interactive display at the Future European Technologies conference in Budapest last week visualized ‘clusters’ of information on the Internet. For example, when looking at the French newspaper Le Monde, articles about similar topics cluster together (see image). The analysis behind the clustering effect looks at the entire article and then determines it similarity by how many words and strings of words are identical. At Le Monde, articles quickly cluster – around topics such as Osama bin Laden’s death – showing the timeliness of news.
Louçã has also visualized clusters within scientific papers. Looking at computer science papers published on the online repository ArXive.org, these cluster much more slowly than newspaper articles. But this is where the technology could be particularly useful for scientists: “If you look at the last two months, or maybe more, you can be aware of emerging scientific trends as you start to see new clusters,” Louçã said.
That’s just one application, however. What Louçã really has his sights set on is being able to map opinion dynamics from newspapers, blogs and social media and then compare it to financial markets to see if there is a correlation.