Sixième séance de 17 juin, 2015

Lieu: STL, Université de Lille 3, Salle Corbin (Bât. B, salle B1 661)

Nous avons le plaisir de vous annoncer la quatrième séance du séminaire « Interactions entre informatique, logique et langage: histoire et philosophie », Mercredi 17 Juin 2015 à 14h, STL, Lille 3, Salle D. Corbin (Bât. B, salle B1 661) avec la participation de Mark Priestley (University College London) et Maarten Bullynck (Sphère, Université Paris Diderot/Université de Paris 8 )

Mark Priestley (University College London),  « Visualizing computation »: from the differential analyzer to the flow diagram, by way of some little known ENIAC drawings.

The first automatic digital computers embodied a rather simple model of the structure of computations as, essentially, repeatable sequences of basic operations. By 1947, however, the structures of arbitrarily complex computations were being routinely described in the new notation of flow diagrams. How, where and why did this more complex model and its associated notation emerge? In this talk I will argue that the ENIAC project played a central role in shaping this new understanding of computational structure, supporting my argument with a look at some little-known diagrams drawn by John Mauchly in 1944.

Maarten Bullynck (Sphère, Université Paris Diderot/Université de Paris 8 ), Excavating the roots of the Chomsky-hierarchy: Computational and linguistic practices at MIT before 1963.

The so-called Chomsky-hierarchy first appeared in an article from 1956 “Three Models for the description of language” and was later clarified in a 1959-article « On certain formal properties of grammars ». The context of the papers is, however, quite different.
Between 1956 and 1959 the research target moves more than once and the tools to catch the target change accordingly. The initial target was to disprove that natural language could be adequately represented as a Markov source, and the device to do that were counter-examples current in discourse analysis. The importance of the finite-state model (that was equivalent to a Markovian source) spurred more detailed research into the limits of what was possible with such a model. Actual implementations of translation or language analysis models at Penn and at MIT showed that some provision for memory should be added. This eventually led to the distinction between context-free and context-sensitive grammars, essentially two models with varying degrees of available memory. Here, the rapprochement with mathematics with the help of the Bourbaki-trained MP Schützenberger was essential to provide all the equivalences between various proposed models. This sequence of events finally leads to the so-called Chomsky hierarchy that would prove of importance for laying a theoretical basis for programming languages.