We propose and implement a methodology for data collection and analysis of Twitter discussions linked to the Argentine economy. Starting with a list of “seed users” later expanded based on following-follower relationships, we build a network of interactions and fetch their tweet timelines. Then, we use a community detection model to compress the structure of underlying relationships and a standard topic model to represent the latent issues discussed in each community. Results suggest that this strategy is able to learn a useful organization and to summarize the contents of social media exchanges of the Argentine economic tweetosphere. Potential applications could be to characterize the links between different economic sectors and to construct community-level indicators of opinions.Sociedad Argentina de Informática e Investigación Operativa