Explaining Grammatical Competence and Performance in an Integrated Connectionist/Symbolic Computational Architecture


Paul Smolensky

John Hopkins University, USA

 

 

Lecture 1. Background & Overview

 

What is cognitive architecture?

                Connectionism and the symbolic architecture.

What is theoretical linguistics?

                Generative linguistics, linguistic universals, and typology.

What is the relation between connectionism and

                symbolic computation?

                generative linguistics?

ICS: The Integrated Connectionist/Symbolic Cognitive Architecture

                Vertically integrated vs. hybrid architectures

Main claim: Attending to neural realization strengthens the theory of universal grammar

 

• Reading:

THM = Smolensky, Paul, & Legendre, Ge'raldine. 2006. The Harmonic Mind: From neural computation to Optimality-Theoretic grammar. Vol. 1: Cognitive architecture; vol. 2: Linguistic and philosophical implications. Cambridge, MA: MIT Press.

Smolensky, Paul & Legendre, Ge'raldine. 2006. Harmony optimization and the computational architecture of the mind/brain. In THM. Chapter 1. Vol. 1, 3–61.

 

 

Lecture 2.  Neural, connectionist and symbolic computation in ICS

 

From neural to connectionist computation

                Distributed representations and vectors

From connectionist to symbolic computation

                Tensor product representations and tensorial networks

Explaining cognitive productivity

                Fodor & Pylyshyn (1988)’s critique of connectionist architecture and ICS

 

• Reading:

Smolensky, Paul & Legendre, Ge'raldine. 2006. Formalizing the principles I: Representation and processing in the mind/brain. In THM. Chapter 5. Vol. 1, 147–205.

 

Lecture 3.  Connectionism and grammar

 

Harmony maximization in connectionist networks

                Network dynamics and optimization

                Parallel soft constraint satisfaction

Harmony maximization in grammar

                Grammars of soft constraints: Harmonic Grammar

                Application: Split intransitivity in French syntax

 

• Reading:

Smolensky, Paul & Legendre, Ge'raldine. 2006. Formalizing the principles II: Optimization and grammar. In THM. Chapter 6. Vol. 1, 207–234.

Legendre, Ge'raldine, Smolensky, Paul, & Miyata, Yoshiro. 2006. Harmonic Grammar and its subsymbolic foundations. In THM. Chapter 11. Vol. 1, 417–452.

Smolensky, Paul. 2006. Harmony in linguistic cognition. Cognitive Science 30, 779–801.

 

Lecture 4. Optimality Theory

 

From Harmonic Grammar to Optimality Theory

Syllable structure typology in Optimality Theory

Syllabification in Berber

                in Optimality Theory

                in Harmonic Grammar

                in a connectionist network

Split intransitivity in French syntax revisited in Optimality Theory

The Harmonic Grammar-Optimality Theory relation

 

• Reading:

Prince, Alan. & Smolensky, Paul. 1997. Optimality: From neural networks to universal grammar. Science 275: 1604–1610.

Prince, Alan. & Smolensky, Paul. 2004. Syllable structure typology: The CV theory. Chapter 6 of Optimality Theory: Constraint Interaction in Generative Grammar. Malden, MA: Blackwell. Reprinted in THM, Chapter 13. Vol. 2, 3–25.

Legendre, Ge'raldine, Sorace, Antonella & Smolensky, Paul. 2006. The Optimality Theory — Harmonic Grammar connection. In THM. Chapter 20. Vol. 2, 339–402.

 

Lecture 5.  Innate grammatical knowledge in a Language Acquisition Device?

 

The concept of an abstract genome

Basic syllabification in a connectionist network

                Computation

                Learning

Encoding the network in an abstract genome

 

 

• Reading:

Soderstrom, Melanie, Mathis, Donald W. & Smolensky, Paul. 2006. Abstract genomic encoding of Universal Grammar in Optimality Theory. In THM. Chapter 21. Vol. 1, 403–471.