Université de Cornell (USA) – Automaton Theories of Human Sentence Comprehension.
Vos rendez-vous sont prévus les 8 juin (14h-16h), 15 juin (14h-16h), 22 juin (14h-16h) et 29 juin (14h-16h) Université Paris Diderot – Bâtiment Olympe de Gouges – 75013 Paris – salle 127
Class 1 : automata = grammar + control date : 8 juin 2015
The class opens by sketching one particular philosophical perspective regarding the positioning of linguistics within the broader realm of cognitive science. It introduces the leading idea of the whole lecture series: namely, that automata combine a grammar with control information. These automata constitute cognitive theories of how comprehension happens. Because they are formalized, they can be studied from a variety of mathematical perspectives. We take up two in particular, generalized left-corner parsing and the notion of informed search applying them to the study of ambiguity-resolution, for instance in Garden-Pathing across English, French and Italian (possibly also German).
Recommended reading: chapters 3 and 4 of “Automaton Theories”
Grillo & Costa 2014 doi:10.1016/j.cognition.2014.05.019
Class 2 : Mechanisms that yield frequency effects date : 15 juin 2015
This class applies the automaton-theoretic perspective to frequency effects in language comprehension. The lecture introduces Cognitive Architectures such as Soar, and illustrates a way to embed parsers within them. From this perspective, the applicability of “chunking” as a theory of practice, becomes applicable to language. This offers a mechanistic interpretation of the information-theoretical complexity metric known as surprisal.
recommended reading: chapters 5, 6 and 8 of “Automaton Theories”
Rauzy and Blache 2012 http://www.lpl-aix.fr/~fulltext/5139.pdf
Class 3 : Information-theoretical complexity metrics date : 22 juin 2015
This class takes up information-theoretical complexity metrics in their own right. We show how to use the freely-available Cornell Conditional Probability Calculator to work out what a person would expect at intermediate points in a sentence. We discuss the implications of this formalization as regards theoretical issues under discussion in Adrian Staub’s class.
Chapter 7 of “Automaton Theories”
and Yun et al 2015 doi:10.1007/s10831-014-9126-6
Chen et al 2014 http://mindmodeling.org/cogsci2014/papers/322/
Class 4 : Neural time-course as evidence for automaton theories date : 29 juin 2015 (14h-16h)
This class applies the automaton-theoretic view to brain data, considering the time-course of signals collected using fMRI. To what degree can these be interpreted as evidence for syntactic structure-building?
Hale et al 2015 http://courses.cit.cornell.edu/jth99/hale-et-al-cmcl15.pdf
Pallier et al 2011 http://www.pallier.org/Publications/Pallier.Devauchelle.Dehaene.pnas.2011.pdf
Smolensky and Legendre 2006 chapter 8 and earlier chapters as needed