The present dissertation presents a pioneer study in what concerns the Portuguese Twitter community analysis and a new system to build real-time datasets from Twitter. During the dataset creation phase we built a system to crawl data from Portuguese Twitter users. This system, TwitterEcho, allows a complete and continuous crawling process, robust enough to support the whole Portuguese community on Twitter. Throughout this dissertation we describe the system implementation details, with more incidence on the client part, since it is a distributed system. Additionally, we present the results of studies done with data retrieved by the system. After, we present the influence model proposed in this dissertation. The model was developed based on models described in the literature or used in commercial applications. We detail the processes of gathering the metrics used to influence measurement. This involves the application of data-analysis techniques to the content produced by Twitter users. Finally, we show the results obtained by the model after its application on a dataset created from TwitterEcho. we also present a preliminary validation of the model, and future ways to validate it.