NON-MONOTONIC KNOWLEDGE REPRESENTATION AND REASONING FOR NATURAL DESCRIPTION OF GOTHIC CATHEDRALS by TYLER

NATURAL DESCRIPTION OF GOTHIC CATHEDRALS by TYLER GREGG CARLSON (Under the Direction of Walter D. Potter) ABSTRACT The goal of the Architecture Represented Computationally (ARC) project is to transform user input or scholarly written descriptions of Gothic cathedrals into logical representations, allowing consistency validation, query-answering, and generation of precise descriptions or visualizations. The work of this thesis is the design and implementation of the first major step of this project, the ARC Logic system, which logically represents domain knowledge and performs inference on this knowledge. The ARC Logic system is domain independent; all domain information is entered by users through input methods, allowing the user to add terminology definitions, facts, and complex rules, without knowing Prolog. This system has a non-monotonic knowledge representation and inference engine, so the system can work with uncertain information, and fill in the information not explicitly stated with background knowledge, narrowing the gap between the logical knowledge representation and natural description in Gothic cathedrals and other real-world domains.