A graph unification machine for NL parsing

A simple, novel, and efficient computational model for a graph unification method for NL parsing is presented. We rely on the body of existing research on labeled graph unification for natural language parsing. This model offers several advantages including: simplicity, efficiency, and amenability to a low-level, efficient, and straightforward implementation. A consequence of this is that some earlier considerations with respect to garbage collection and redundant node copying become obsolete. The model uses a novel feature of subnode structure sharing.