Subsymbolic Natural Language Processing

Risto Miikkulainen draws on recent connectionist work in language comprehension tocreate a model that can understand natural language. Using the DISCERN system as an example, hedescribes a general approach to building high-level cognitive models from distributed neuralnetworks and shows how the special properties of such networks are useful in modeling humanperformance. In this approach connectionist networks are not only plausible models of isolatedcognitive phenomena, but also sufficient constituents for complete artificial intelligencesystems.Distributed neural networks have been very successful in modeling isolated cognitivephenomena, but complex high-level behavior has been tractable only with symbolic artificialintelligence techniques. Aiming to bridge this gap, Miikkulainen describes DISCERN, a completenatural language processing system implemented entirely at the subsymbolic level. In DISCERN,distributed neural network models of parsing, generating, reasoning, lexical processing, andepisodic memory are integrated into a single system that learns to read, paraphrase, and answerquestions about stereotypical narratives.Miikkulainen's work, which includes a comprehensive surveyof the connectionist literature related to natural language processing, will prove especiallyvaluable to researchers interested in practical techniques for high-level representation,inferencing, memory modeling, and modular connectionist architectures.Risto Miikkulainen is anAssistant Professor in the Department of Computer Sciences at The University of Texas atAustin.