Information graphs of event threads: Identifying and building forward and backward event threads for societal events

This research presents an approach for creating semantically linked event threads originating from free text sources and for developing an information theory-based mathematical foundation for characterizing the event threads. Events are represented as components mapped into ℝ^N each consisting of sets of attributes and reliabilities of the sources. These components can pertain to an event description, temporal entities, location, person/agent involved, and objects involved in the event. The application of this research is be able to identify a significant societal event and then extract event threads extending both forward and backward in time and link them by their semantic relationships. Information measures based on algebraic topology use simplicial complexes to identify important threads from the less important spin-off threads and disjoint threads or events. Interactions of event threads are discussed.