Mapping the world with social buzz

With the advancement of information technology, social networks have become popular platforms for people to share and discuss their activities and interests. The social network contents, the social buzz in these networks, are a valuable source for us to learn about the world. Meanwhile, the development of geotagging technology allows us to link such contents on these virtual networks with real-world locations. Taking advantage of the popularity of social networks and the geotagging technology, in this project, we desire to map the world with social buzz. In other words, we aim at building a map system enriched by social network contents, which is beneficial to both individuals and businesses, as the area profiles summarized from social buzz can help them to understand the areas conveniently. We build our system in three steps: collecting contents, training the model and presenting the model. The core of our system is a tree-structure model that summarizes social network contents for areas in different levels. In this model, we profile each area's popularity, topic and geographic information. With this model, we develop a generative process to describe how a document is composed. An estimation inference algorithm based on Gibbs sampling is provided to learn an optimal tree model from a set of observed documents. Evaluation results show that our system is effective