Learning an OT grammar with a structured candidate set. Invited talk given at: Linguistic Colloquium, January 10, 2007, Bielefeld, Germany. Slides.
In Optimality Theory (OT, Prince and Smolensky 1993/2004), the surface form is predicted to be the most harmonic (the
optimal, the best) element of a candidate set that is generated from the underlying form. Hence, OT requires an
optimization algorithm that is able to find the best element of the candidate set with respect to the target function
called Harmony. In my dissertation (Bíró, "Finding the Right Words", 2006), I introduced the Simulated Annealing for
Optimality Theory Algorithm (SA-OT), which can solve the problem more or less accurately. Actually, the errors made
by the algorithm are interpreted as performance errors, and therefore SA-OT is argued to be a cognitively plausible
model for linguistic performance.
The SA-OT Algorithm requires the candidate set to be structured by some topology (neighborhood structure). Although
divergent from standard OT, this concept is not unknown in heterodox versions of Optimality Theory, for instance in
Persistent OT ("harmonic serialism") by McCarthy (2006). Following an introduction to OT and to the SA-OT Algorithm,
the goal of the talk will be to analyze the role of a neighborhood structure in OT, and especially its implications
to learnability. The results of some preliminary learning experiments
will also be presented.