|
Harmonic Explanation in the theory of linguistic performance| old_uid | 5379 |
|---|
| title | Harmonic Explanation in the theory of linguistic performance |
|---|
| start_date | 2008/10/14 |
|---|
| schedule | 10h |
|---|
| online | no |
|---|
| details | séminaire double |
|---|
| summary | “Harmonic Explanation” is a strategy for explaining language phenomena as the result of optimization of Harmony, a measure of the well-formedness of representations. Harmonic Explanation originated in neural network theory as a means of understanding the emergent properties of network computation at a level higher than that of individual units and connections; it then became the central mode of explanation in a grammar formalism, Optimality Theory, originally developed within phonology (Prince & Smolensky 1993/2004) but since extended to syntax (Legendre, Grimshaw & Vikner, 2001) as well as semantics and pragmatics. The work presented in the talk is the beginning of an attempt to bring Harmonic Explanation, through Optimality Theory, back to neural network modeling of linguistic behavior, in particular, phonological production. Network (Lotka-Volterra) attractor dynamics for selecting a discrete output combine with stochastic (Hybrid Monte Carlo) dynamics for optimizing Harmony to permit, in simple cases, Harmonic Explanations of general error patterns and reaction times. The goal is a unified explanatory framework for linguistic competence, performance, and statistics. |
|---|
| responsibles | Faraco, Bertrand |
|---|
| |
|