Abstract:
Since "The Harmonic Mind" of Paul Smolensky and Geraldine Legendre appeared
in 2006, there has been an increased interest in Harmony Grammar (HG) and in
Simulated Annealing (SA) within the Optimality Theoretical (OT) camp. Yet, many
phonologists might find it difficult to understand the mathematical and
computational details of this huge work. To tell the truth, they even do not care
about the connectionist component of Optimality Theory. It was also in 2006 that I
defended my thesis (Finding the Right Words: Implementing Optimality Theory
with Simulated Annealing, phd-thesis. GroDiL 62; ROA-896), in which I had
developed a slightly different approach, called the Simulated Annealing for
Optimality Theory (SA-OT) Algorithm. Mine, let me argue, has the following three
practical advantages compared to the approach of Smolensky and al: (1) it is purely
symbolic, and not connectionist; (2) it is really OT, and not HG; (3) it has been
demonstrated to model linguistic performance, including the influence of speech rate
on performance errors.
In this talk, I will try to give a short introduction to OT, HG, SA and SA-OT so
that the recent literature in "more computationally oriented" OT become more
accessible to "less computationally minded" linguists. To conclude the talk, a piece
of software will also be presented where one can "play around" with an SA-OT
model.