Work

"Generalize and Sift" as a Model of Inflection Acquisition

Year: 2001
Type: dissertation
Abstract: In this thesis, I propose the "Generalize and Sift" model for the acquisition of inflection in natural language.This model uses a two step learning process along with probabilistic selection when there is uncertainty.To learn, the model first creates rules in a specific to general search over the phonemic feature space of words, allowing for a fast first degree approximation to the language's rules.Then, in cases where rules overlap, the model weights the applicable rules and adjusts these weights according to observed data, eventually converging on the correct rule.I have implemented this model in a computer program and run simulations with data mimicking the words a child might hear as he learns inflections.The implementation displays behavior very similar to those of children: the same learning curve, similar mistakes while learning, and equivalent behavior once it reaches "adulthood.
Language: English
Field: Psychology
Open Access status: green
OpenAlex ID: W2281297002
Cites: 26
Cited by: 38
Related to: 10