Model doesn't argue against rule-based explanation
The past-tense model is problematic in numerous ways—eg it can’t represent certain words, it can’t learn certain rules, it learns rules found in no human language, and it fails at its assigned task of mastering the past tense of English.
To overcome such difficulties, connectionism will have to implement certain features of rule-based, symbolic series.

So, connectionist models are either inadequate as a model of language, or at best offer an implementation of classical rule-based accounts.

Stephen Pinker and Alan Prince.
 
Note: Compare the structure of this argument to that of "The Connectionist Dilemma", Box 31.
PAGE NAVIGATOR(Help)
-
Artificial Intelligence »Artificial Intelligence
Can computers think? [1] »Can computers think? [1]
Yes: connectionist networks can think [5a] »Yes: connectionist networks can think [5a]
Connectionist networks can think without following rules »Connectionist networks can think without following rules
The Past-Tense Acquisition Model »The Past-Tense Acquisition Model
Model doesn't argue against rule-based explanation
Any system can be described by a dual-route model  »Any system can be described by a dual-route model
The Connectionist Dilemma »The Connectionist Dilemma
+Commentaar (0)
+Citaten (0)
+About