chapter  5
14 Pages

Inside the Path Evaluator

In C hapter 3 we discussed two different approachs to markerpassing. The first, based on Quillian’s model, uses marker-passing as a “ dum b” mechanism th a t returns all the paths th a t can be found.2 The second, more like Fahlm an’s scheme, makes markerpassing “sm arter” — the goal is for only correct paths to be reported. The mechanism in SCRAPS, and the mechanism I advocate here, is the former.