TWO KINDS OF PHONOTACTICS: ABSOLUTE AND COMPARATIVE

What is the phonological well-formedness of a particular word? How is it learned? Hayes and Wilson (2008) suggested both a grammar framework and a learning system. Framework: the maxent variant (Goldwater and Johnson 2003) of harmonic grammar (Legendre et al. 1990), with the overall constraint-based architecture borrowed from Optimality Theory (Prince and Smolensky 1993) Hayes, CLS April 2014 Comparative Phonotactics p. 4 The system assigns a predicted probability to every word; difference in probability correspond to differences in well-formedness. An algorithm selects phonological constraints and weights them so as to maximize the predicted probability of the set of existing words. To some extent, the output probabilities assigned by this system succeed in matching linguists’ phonotactic descriptions and human phonotactic intuitions. Hayes, CLS April 2014 Comparative Phonotactics p.