Context variability promotes generalization in reading aloud: Insight from a neural network simulation.
Year: 2020
Type: conference-paper
Abstract: How do neural network models of quasiregular domains learn to represent knowledge that varies in its consistency with the domain, and generalize this knowledge appropriately?Recent work focusing on spelling-to-sound correspondences in English proposes that a graded "warping" mechanism determines the extent to which the pronunciation of a newly learned word should generalize to its orthographic neighbors.We explored the micro-structure of this proposal by training a network to pronounce new made-up words that were consistent with the dominant pronunciation (regulars), were comprised of a completely unfamiliar pronunciation (exceptions), or were consistent with a subordinate pronunciation in English (ambiguous).Crucially, by training the same spelling-to-sound mapping with either one or multiple items, we tested whether variation in adjacent, within-item context made a given pronunciation more able to generalize.This is exactly what we found.Context variability, therefore, appears to act as a modulator of the warping in quasiregular domains.
Source: TSpace (University of Toronto)
Authors: Ian S. Miller, Nicolas Dumay, Mark A. Pitt, Brian Trung Lam, Blair C. Armstrong
Language: English
Subfield: Developmental and Educational Psychology
Field: Psychology
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W3135078845