Event detection, event characterisation and community detection on evolving networks

Abstract

Since its inception the internet has revolutionized the way we communicate and interact with each other, how we acquire and share information as well as how we socialize and form communities. Social media platforms have significantly contributed to this revolution with billions of individu- als using them on a daily basis all over the world. Anyone can ask a question, share personal information, opinions, media, news or content generated from other users and immediately reach any individual around the globe, forming an ever evolving network of interactions and relations. This makes clear that social and online media can be a valuable source of information and could help us reveal trends in the news and structures on them using techniques such as data mining and social network analysis. Also social and online media diffuse information instantly in contrast with traditional media that take much longer to make available any information, we can benefit from it by developing news detection methodologies that use streams of social or online media and monitor their evolution over time. In this work the goal is to utilize social network analysis on evolving networks for discovering latent structures on online communities and conceive methodologies that extract useful information from online sources such as news articles, posts on social media and forums. This work has demonstrated what we can achieve by applying social network analysis on evolving networks. In chapter 3 we were able to develop a novel event detection methodology that outperformed state of the art approaches and provided further insights such as a comprehensive summary of the event and the sentiment and emotions of users that were discussing about it. The method was applied in a number of heterogeneous data sets consisting of news articles and Twitter posts, for evaluating it and demonstrating how it works. We also used this methodology to detect the main events of the COVID-19 pandemic as it has been discussed on Twitter compiling a retrospective collection of viral events. In chapter 4 by combining social network analysis and evolving networks we revealed com- munities of software developers on the Stack Overflow platform based on what technologies they ask and answer questions. This work shows what we can achieve by applying social network analysis on evolving networks and provides two use cases that demonstrate this, pointing out that evolving networks should be one of the fields that social network analysts should focus on for further research in the future.