Lieu: MESHS, 2 Rue des Canonniers, 59000 Lille
Nous avons le plaisir de vous annoncer la prochaîne séance du séminaire « Histoire et Philosophie de l’Informatique”, mercredi 5 avril 2017 à 14h, MESHS, Lille. Le thème de cette séance est: « Informatique et linguistique » avec la participation de: Geoffrey Pullum et Christopher Pinon.
Geoffrey Pullum (University of Edinburgh), Emil Post, computably enumerable sets, and the description of human languages
Emil Post’s « production systems » are ancestral to the nondeterministic random string enumeration procedures that linguists call generative grammars. Post employed them as tools for studying (i) completeness, consistency, and decidability of logics, and (ii) computably enumerable (c.e.) sets of numbers. Post coded positive integers as finite strings over a singleton alphabet. But the symbol inventory of a production system can be of arbitrary finite size: it could be the set of words in the dictionary. Production systems could therefore be employed to define sets of strings of words in a human language. Linguists have developed this idea for more than 60 years. But generative grammars are in some ways strikingly ill-suited to the task of characterising human languages. Post was interested in the structural properties of sets, not their elements. Linguists’ interests are the opposite: the structure of clauses and phrases is what matters, not the structure of infinite sets of them. The radical implications of this point have been unwisely overlooked. Formal language theory is standardly assumed to be the appropriate mathematical basis for linguistics, but I argue that this assumption should be relinquished; linguists should instead look to descriptive complexity and finite model theory.
Christopher Pinon (UMR 8163 STL, Université de Lille), Subcategorization in early generative grammar