Abstract
In this work, measures of information diffusion across the community structure of Internet blogs are examined for early dynamic characterization potential of events that may lead to real-world violence. Social media topology is observed through focused crawling of web pages and blogs related to incidents of provoked social upheaval, and our introduced Community Entropy Ratio (CER) is used as a proxy for information spread. We introduce a blog topology that contains both explicit relationship via hyperlinks and implicit relationships via lexically relevant author co-comments. We then show how CER performs on two real-world topologies by tracking dispersion over time of various lexical terms. We then compare the CER results with the known truth for the timelines of the events and show where the metric succeeds and fails in providing an initial indicator. While there are some data volume concerns, the metric correctly identifies one of the case studies as having a high potential for eventual violence.