The topics of this thesis are event detection and social network analysis in social media. Our work centres on Geo-tagged User Generated Content (UGC) in Twitter, such as Twitter data generated from the metropolitan area of Dublin Ireland over a one month period of time. In this thesis we address the problem of how to detect small scale unexpected events using UGC both in real-time and retrospectively. We proposed a language-text joint modeling algorithm to cope with the large volume and unstructured nature of UGC. We also demonstrate our discovery of interesting correlations between a Twitter user’s social communities and their mobility patterns. Finally a set of features are proposed for carrying out Twitter user’s account type classification, for the purpose of irrelevant contents filtering. This thesis includes several experimental evaluations using real data from users and shows the performance of our algorithms in event detection and provide evidence for our discoveries.