A Comparison between Ot and Hg from a Computational Perspective

Optimality Theory (OT) and Harmonic Grammar (HG) differ because the former assumes a model of constraint interaction based on strict domination, while the latter assumes a weighted model of interaction. As Prince and Smolensky (1997) admit, “that strict domination governs grammatical constraint interaction is not currently explained”. Yet, Legendre et al. (2006, 911-912) make two suggestions. The first suggestion is that OT’s strict domination might have algorithmic advantages, in the sense that it “may enable quick-and-dirty optimization algorithms [. . . ] to consistently find a single global [. . . ] optimum, whereas arbitrarily weighted constraints typically lead such algorithms to produce widely varying solutions, each only a local optimum.” The second suggestion is that OT’s strict domination might have learnability advantages: “another possibility is that demands of learnability provide a pressure for strict domination among constraints”, although they note that “it remains an open problem to formally characterize exactly what is essential about strict domination to guarantee efficient learning.”