Some theoretical and experimental observerations on naive discriminative learning (Updated version)

Abstract

Natural language use is full of choices among multiple possible alternatives, whether phones, words, or constructions, which are influenced by a large number of contextual factors, and which rather exhibit asymptotic, imperfect tendencies favoring one or more of the alternatives, instead of single, categorical, perfect choices. This contrasts with item-by-item learning in simple controlled experiments which typically have been modelled by the Rescorla-Wagner equations. We find the former "messy" types of problems as a key area of interest in modeling and understanding language use, and consequently consider the application of the Rescorla-Wagner equations in the form of a Naive Discriminative Learning classifier to such complex phenomena of considerable utility in linguistic research.

This is an updated version. (Previous version: http://dx.doi.org/10.15496/publikation-8620)