Temporal Methods to Detect Content-Based Anomalies in Social Media

We develop a method for time-dependent topic tracking and meme trending in social media. Our objective is to identify time periods whose content differs significantly from normal, and we utilize two techniques to do so. The first is an information-theoretic analysis of the distributions of terms emitted during different periods of time. In the second, we cluster documents from each time period and analyze the tightness of each clustering. We also discuss a method of combining the scores created by each technique, and we provide ample empirical analysis of our methodology on various Twitter datasets.