Harmonic Grammar and Harmonic Serialism - John J. McCarthy

Harmonic Grammar and Harmonic Serialism - John J. McCarthy

12. Learning Serial Constraint-based Grammars

Harmonic Grammar and Harmonic Serialism - John J. McCarthy

Robert Staubs [+-]
Independent scholar
Robert Staubs received his PhD in 2014 from the University of Massachusetts Amherst. He specializes in phonological theory, learning, and models of typological frequency


This paper proposes a method for learning grammars in the general framework of Harmonic Serialism, a variant of Optimality Theory with gradual derivations. It addresses problems of structural ambiguity introduced by derivations, and is able to deal with variable as well as categorical data. Maximum Entropy serial grammars generate an unbounded number of possible derivations. By formalizing the derivational space in terms of Markov chains, probabilities over those derivations can be calculated, allowing the use of standard optimization methods for fitting the grammar model to the learning data. Learning results are provided for simplified versions of an opaque stress-epenthesis interaction, and of a variable vowel deletion process modeled after French schwa.

Notify A Colleague


Staubs, Robert. 12. Learning Serial Constraint-based Grammars. Harmonic Grammar and Harmonic Serialism. Equinox eBooks Publishing, United Kingdom. p. 369-388 Sep 2016. ISBN 9781845531492. https://www.equinoxpub.com/home/view-chapter/?id=24953. Date accessed: 23 Mar 2018 doi: 10.1558/equinox.24953. Sep 2016

Dublin Core Metadata