Modeling Microblogs using Topic Models

As the popularity of micro-blogging increases, managing friends and followers and their tweets is becoming increasingly complex. In this project, we explore the usage of topic models in understanding both text and links in micro-blogs. On a data set of 21306 users, we find that LDA can find good topics that seem to capture meaningful topics of discussion in twitter. We also find that knowing whether two users tweet about similar topics is more useful in predicting links between them than standard network analysis metrics that ignore text.