Data-driven characterization of viral events on social networks: sustainability issues in the palm oil production

Palm oil is the most widely used vegetable oil in the world. Yet, its production and consumption have generated heated debates over the past few decades due to its environmental impact. In this thesis, we study the debate on palm oil on the popular social network Twitter since 2006. Using Natural Language Processing and Network Science tools, we analyze the most important viral events related to palm oil. We identify the role that Opinion Drivers play in the debate and how most debates are short-lived. Indeed, by studying the interevent time distributions of certain hashtags, we see that even the most far-reaching viral events are quickly forgotten. Furthermore, most viral events are described by similar characteristics, showing an underlying universality that goes beyond the specific topics. All in all, our results show that the public debate on Twitter is limited to a few countries and mainly centered around the leading actors of public opinion. Thus, rather than considering this debate as something intrinsic to the public, it should be regarded as something mainly driven by a few organizations.