Script-based Reasoning for Situation Monitoring

An expert system that monitors complex activity requires knowledge that is difficult to capture with standard rule-based representations. The focus of this research has been to design and implement script-based reasoning techniques integrated into a rule-based expert system for situation monitoring to address this problem. The resulting expert system, SCripted ANalyst (SCAN), for battlefield monitoring has the capability of reasoning about tactical situations as they develop and providing plausible explanations of activities as inferred from intelligence reports. Sequences of events are monitored through the use of script templates which are matched against events and the time relations between events. SCAN detects causal relations between events, generates multiple hypotheses, fills in information gaps, and sets up expectations about time-dependent events-all features a simple rule-based expert system cannot easily provide.