Graphical Deep Knowledge for Intelligent Machine Drafting

Abstract

The problem of Intelligent Machine Drafting is presented, and a description of an existing implementation as part of a graphical generator function is given. The concept of Graphical Deep Knowledge is defined as a representational basis for Intelligent Machine Drafting problems as well as for physical object displays. A (partial) task domain analysis for Graphical Deep Knowledge is presented. Primitives that are necessary to deal with a world of 2-D forms and colors are introduced. Among them are primitives for describing forms, positions, parts, attributes, sub-assemblies, and an abstraction hierarchy. The use of the linearity principle for knowledge structure derivation from natural language utterances is shown.