Connectionist approaches to generative phonology

John Alderete, Paul Tupper

Abstract:

While connectionist models are ubiquitous in psycholinguistic approaches to language processing, they are less well-known as generative models of grammar. This work surveys a literature in which connectionist models have been developed to address problems central to generative phonology. The focus is on explaining to the newcomer how precisely these models work, and how they grapple with locality, gradience, opacity, and learnability in phonology. An understanding of connectionist phonology gives both a deeper understanding of past developments in phonological theory and a glimpse into its future. 

Keywords: connectionism, neural networks, exemplar phonology, locality, gradience, opacity, learnability

Downloads:  weblink to PDF

Full citation: Alderete, John and Paul Tupper. 2016/To appear. Connectionist approaches to generative phonology. To appear in Anna Bosch and S. J. Hannahs (eds.), The Routledge handbook of phonological theory.