Integrated Framework for Knowledge Representation and Acquisition

摘要

The goal of the present project is to build a knowledge-representation framework which is sufficiently powerful and flexible to support the tasks of understanding events, problem-solving, and knowledge acquisition. This is accomplished by integrating several major cognitive psychology and artificial intelligence (AI) frameworks: schemas serve as the general organizers of expectancy-driven knowledge; production rules represent situation specific details; and mental models represent causal relationships that underlie events. The integrated framework was implemented and tested on a small-scale AI system, INKS (for INtegrated Knowledge System), which reads a story in parsed format, finds a planning failure, tries to determine the causes of the failure, recommends fixes based on the causal analysis, and updates its memory when it finds a successful solution. Ways of expanding the framework and INKS are discussed.