News & Events

Computational Locality and Morpho-phonological Learning

Jane Chandlee

Haverford College

Bausch and Lomb 106

Computational models of phonological grammar learning (including a lexicon of abstract underlying representations, orURs, and a mapping from those URs to surface representations, or SRs) have long been a topic of interest, with significant progress made in theoretical frameworks that assume an innateor provided constraint set that the learner must rank or weight(e.g., Jarosz 2006, Tesar and Prince 2007, Apoussidou 2007,Pater et al. 2012, Tesar 2014, among many others). In this talk I will present a model that instead capitalizes on proposed formal language theoretic properties of these grammars—namely strictlocality (Chandlee and Heinz 2018)—which serve to narrowthe hypothesis space of grammars such that a learner canconverge with a relatively small amount of data. The implications and applications of this model will also be discussed in two areas: our understanding of phonological acquisition and the development of technology for low-resource languages.