A SOFTWARE ARCHITECTURE FOR DEVELOPING AGGREGATE INTELLIGENCE SUMMARIES

WHILE INTELLIGENCE INFORMATION IS CRUCIAL TO ALL DECISION ANALYSIS AND OPERATIONS EVALUATION, ITS COLLECTION AND ANALYSIS HAS BECOME A DEMANDING TASK. RECENT YEARS HAVE SEEN AN INCREASE IN THE AMOUNT AND SCOPE OF FIRST LEVEL INTELLIGENCE PRODUCTS WHILE THE TIME AVAILABLE FOR ANALYSIS AND DECISION MAKING HAS DECREASED. A COMPUTERIZED SYSTEM CAPABLE OF CREATING ANALYTICAL INTELLIGENCE REPORTS FROM FIRST LEVEL INTELLIGENCE PRODUCTS WOULD GREATLY ASSIST DECISION MAKERS. PAST EFFORTS TO CREATE COMPUTERIZED SYSTEMS HAVE BEEN HINDERED BY THE DIFFICULTY IN EVALUATING INTELLIGENCE HYPOTHESES IN THE LIGHT OF UNCERTAIN OR MISSING EVIDENCE. MRJ AND GEORGE MASON UNIVERSITY (GMU) PROPOSE AN INNOVATIVE SYSTEM THAT COMBINES STATE OF THE ART PROBABILISTIC INFERENCE TECHNIQUES WITH SYMBOLIC KNOWLEDGE-BASED TECHNIQUES. PROBABILISTIC INFERENCE METHODS HAVE BEEN SHOWN TO BE WELL SUITED TO EVALUATING HYPOTHESES WITH UNCERTAIN OR MISSING EVIDENCE. RECENT RESEARCH HAS DEMONSTRATED THAT KNOWLEDGE-BASED TECHNIQUES CAN BE USED TO CREATE THE PROBABILISTIC MODELS NECESSARY FOR PROBABILISTIC METHODS TO PERFORM THEIR ANALYSIS. A SYSTEM COMBINING THESE TWO TECHNOLOGIES COULD AGGREGATE FIRST LEVEL INTELLIGENCE TO DEVELOP AND TEST PLAUSIBLE INTELLIGENCE HYPOTHESES AND CREATE ANALYTICAL INTELLIGENCE REPORTS. THE RESULTING SYSTEM WILL GREATLY ENHANCE OUR ABILITY TO ANALYZE FIRST LEVEL INTELLIGENCE PRODUCTS. THE SYSTEM DESCRIBED WOULD LOWER THE HUMAN LABOR REQUIRED TO COLLECT FIRST LEVEL INTELLIGENCE AND GENERATE SUMMARIES. IT WOULD MAKE THESE REPORTS MORE AVAILABLE TO DECISION MAKERS, ESPECIALLY IN TIME CRITICAL SITUATIONS. THIS SYSTEM COULD EITHER LOWER THE COSTS OF WAR GAME SIMULATIONS OR THE COST SAVINGS COULD BE REDIRECTED INTO CREATING MORE DETAILED SIMULATIONS. THE BASIC TECHNOLOGY COULD BE SUPPLIED DIFFERENT DOMAIN KNOWLEDGE TO CREATE COMMERCIAL APPLICATIONS.