Structured Probabilistic Attributes, Relationships, and Knowledge (SPARK)

Modern conflict is complex and dynamic. Although US forces"ability to collect data has risen to meet the challenges posed by this dynamic complexity, vast stores of information are not, by themselves, sufficient to support an effective Observation, Orientation, Decision, and Action (OODA) cycle. Without methods of modeling the individuals and groups in the battlespace, and detecting and understanding meaningful change in the patterns of activity, Warfighters will suffer from impeded and incomplete understanding of the social context of the conflict. The Structured Probabilistic Attributes, Relationships and Knowledge (SPARK) project directly addresses this problem through four technical thrusts: SPARK"s social network construction capability is able to distill detailed, domain-specific social network models from large corpora of unstructured text. A logic-based probabilistic modeling and inference capability allows Warfighters and analysts to use complex reasoning processes to analyze large network models. SPARK"s temporal modeling capability exposes the patterns and trends of network behavior and structure, and helps analysts isolate the signal of long-term change from the noise of ordinary short-term activity. Finally, an intuitive and flexible user interface based on the Ozone widget framework will be developed and evaluated in the operational setting of the INSCOM Futures Lab.