SOCIAL MEDIA FOR THE SUPPORT AND IMPROVEMENT OF CITIZENS' WELL-BEING

Social Media (SM) platforms are a widespread phenomenon connected to the diffusion of the Internet. This work will exploit several SM platforms belonging to different categories to prove how collective intelligence, SM mining and data analysis can be used to improve and support citizens safety and health. Consequently, this thesis proposes a set of multidisciplinary, cost-effective approaches to investigate and sustain public health and well-being through SM. For example, SM platforms are becoming increasingly crucial in providing uncensored information on drug sourcing mechanisms or, in case of emergency, revealing their usefulness to support population and to get more situational awareness. In fact, in recent years, a significant increment of SM activity has been observed in the aftermath of mass convergence and emergency events. To this regard, microblogs such as Twitter, Weibo, and Instagram are favored channels of information diffusion because of their ubiquity and simplicity. During emergencies, people usually report their experience on these media, which are consequently overwhelmed by information concerning the unfolding scenario. This SM feature is explored in this work through a system able to process incoming data to identify useful information for these purposes, analyzing the data using a two-fold perspective. On the one hand, we explore damage detection techniques to detect messages reporting damage to infrastructures or injuries to the population. On the other hand, we propose a message geolocation component that performs the geoparsing task by exploiting online semantic annotators and collaborative knowledge-bases. Furthermore, this work measures the relevance of drugs diffusion and advertisement, as well as user engagement using SM and proposes a semi-supervised approach to support health departments in identifying novel substances timely. Unfortunately, SM platforms are also the ideal plaza for the proliferation of other harmful information. Cyberbullying, sexual predation, self-harm practices incitement are some of the effective results of the dissemination of malicious information on SM. The hate can be directed towards wide groups of individuals, discriminated for some features, like race or gender. To study and monitor the phenomenon of hate in SM we propose a methodology to prevent the critical social consequences of massive online hate campaigns and will compare the approach with the results of other academic works. Finally, concerning these kinds of attacks and news spreading, we will discuss the problem of censored identities and propose a methodology to spot them.