Mind and Language, 9, 273-287, 1994.


Generalization and Connectionist Language Learning



Morten H. Christiansen & Nick Chater


Abstract

The performance of a learning system may be assessed by its ability to generalize from past experience to novel stimuli. Recently, Hadley (1994) has criticized connectionist language learning researchers for not approaching generalization in a sophisticated manner. This paper discusses Hadley's thesis concerning generalization and connectionist language learning. Motivated by linguistic considerations, we provide more formal and precise definitions of three varities of generalization. Connectionist simulations are presented using simple recurrent networks. The simulation results are evaluated in the light of the revised definitions, and it is concluded that connectionist models can, at least in part, accommodate sophisticated generalization. We consider the prospects for connectionist and other approaches to language learning for meeting these criteria in general.
Click here to download a PDF version.

rule.gif (155 bytes)

Home | People | Research | Links | Contact | Publications | Presentations
Cognitive Neuroscience Laboratory

Please email suggestions/errors to mhc27@cornell.edu