As modern deep networks become more complex, and get closer to human-like capabilities in certain domains, the question arises as to how the representations and decision rules they learn compare to the ones in humans. In this work, we study representations of sentences in one such artificial system for natural language processing. We first present…

Examines processes by which general but weak search methods are transformed into powerful, domain-specific search strategies by classifying types of heuristics learning that can occur and components that contribute to such learning. A learning system--SAGE.2--and its structure, behavior in different domains, and future directions are explored. (36…