Monte Carlo Simulation for Plausible Interpretation of Natural-Language Spatial Descriptions

This system generates plausible threedimensional visualizations of basic English descriptions of a zoo environment (mainly animals and plants) by means of a Monte Carlo simulation. It combines a semantic network and an inheritance-based knowledge base as representations for explicit and implicit spatial information, respectively. Its linguistically motivated aspects address underspecification, vagueness, uncertainty, and context, as well as intrinsic and deictic frames of spatial reference. The underlying reasoning formalism is probability-based geometric fields, which are used for qualitative constraint satisfaction.