Knowledge-Based Constraint Satisfaction for Spatial Reasoning

This linguistically motivated work addresses issues in reasoning intelligently over spatial descriptions of simple, static scenes to produce plausible graphical interpretations. It uses a combined representation that couples a semantic network for explicit knowledge with a knowledge base of frames for implicit knowledge. The knowledge base contains generalized rules for interpreting what objects are and how they should and should not be interpreted alone and in spatial interrelationships. The linguistic emphasis is on the semantics and pragmatics of underspecification, vagueness, uncertainty, and context in reasoning over spatial language and knowledge.