Development of an Advanced Method for the Analysis of Topics and Events on Twitter and their Evolution

Twitter has evolved in recent years from a social network used to exchange opinions among friends to a platform for sharing information about events and trending topics; therefore Twitter, thanks to its powerful APIs, can be used a huge public source for the analysis of events, topics and their evolution over time. Twitter presents its content to users as a raw stream of chronologically ordered tweets; this makes it extremely difficult to identify interesting patterns and trends. Many researchers have tried to overcome this problem by focusing on techniques for the analysis and visualization of tweets in a great variety of fields, from politics to natural disasters. Unfortunately most of these works are either focused on very specific topics and/or based on an offline analysis of previously collected tweets. The aim of this thesis is to develop a method for the analysis and visualization of events and topics on Twitter and their evolution over time. This method can be applied to any unconstrained dataset of tweets collected from the Twitter API and is based on four main components: content analysis of tweets, pattern analysis of mentions and retweets, identification of opinion makers and influential tweets, and sentiment analysis.