ABSTRACT
1 Symbolic and distributed connectionist systems live in distinct processing niches. Symbolic systems supply variables, bindings, logical rules, constituent structure, tokens (versus types), hierarchies and inheritance. In contrast, distributed connectionist systems supply statistically based associations, reconstructive memories, graceful error degradation, automatic category and prototype formation, and generalization to novel instances. Clearly, a synthesis is desirable and this chapter describes a technique, called symbol recirculation, for automatically forming lexical representations as patterns of activation distributed over ensembles of connectionist units. To show the potential for natural language processing with these distributed symbols, two script-based natural language comprehension and para-phrasing systems are briefly described, each using lexical entries formed through symbol recirculation.
