Abstract
An architecture of a knowledge-based system is presented for decision support in sociopolitical domain. Human experts' decisionmaking is characterized by their efficient heuristics: simplifying the world reality getting the idea on what kinds of events naturally belong together, and grasping what sorts of behavioral patterns appear in sequences of observed events. In this paper, the system is designed that can generate dynamic evolving patterns of events with different abstraction levels using meta-knowledge, universally accepted as human repetitive patternized social behaviours, above the empirical knowledge which is acquired from documents as causally-chained networks by our suggesting knowledge representation formalism. Those generated event patterns at each level evolve in parallel concurrently under the domination of a social behavioral plan-scheme of the upper level. Based on such a human memory-like knowledge organization, the system enables decision-makers' flexible and efficient access to the knowledge store on their individual demands, as well as provides them with comprehensive information from global viewpoints, both of which contribute to explicating the problems in their pre-decision stages.
