Can the EU's Sensei project predict Brexit by data-mining social media chatter?

The Sensei project, backed by the European Commission, is analyzing millions of online conversations to predict events, such as Brexit: the UK vote on whether to retain its EU membership. Barcelona-based Websays, one of the project's members, has already claimed a "stunning success" in predicting the Spanish general election. Websays' founder, Dr Hugo Zaragoza, said in a statement (PDF): "There are others out there listening to social media, but we are the only ones with this deep combination of technology and human analysis through machine learning. We think this can add real value to business and help us to predict outcomes better than anyone else." The hope is that "sentiment analysis" will prove more accurate than traditional polling systems.