[Prev][Next][Index]

CSLI Calendar, Oct. 15, 3:3


       C S L I   C A L E N D A R   O F   P U B L I C   E V E N T S
_____________________________________________________________________________
15 October 1987                    Stanford                      Vol. 3, No. 3
_____________________________________________________________________________

     A weekly publication of The Center for the Study of Language and
     Information, Ventura Hall, Stanford University, Stanford, CA 94305
                              ____________
           CSLI ACTIVITIES FOR THIS THURSDAY, 15 October 1987

   2:15 p.m.		CSLI Seminar
     Classroom          The Acquisition of Morphology
     Ventura Trailers  	Discussion of the debate between
			Rumelhart/McClelland and Pinker/Prince  
			Dave Rumelhart, (der@psych.stanford.edu)
			Abstract below
			
   3:30 p.m.		Tea
     Ventura Hall		

   4:15 p.m.		CSLI Colloquium
     Classroom		A Logic for Practical Reasoning
     Ventura Trailers	Tim Flannagan, Logica Cambridge
			Abstract in last week's Calendar

                             --------------
           CSLI ACTIVITIES FOR NEXT THURSDAY, 22 October 1987

   12 noon		TINLunch
     Ventura Hall       Reading: "On Language and Connectionism:
     Conference Room  	Analysis of a Parallel Distributed Processing
			Model of Language Acquisition"
			by Steven Pinker and Alan Prince
			Discussion led by Dave Rumelhart
			(der@psych.stanford.edu)
			This TINLunch is a follow up of the seminar
			that will have been given on 15 October. Please
			see the abstract for it below. 

   2:15 p.m.		CSLI Seminar
     Room G-19          External Language and Internal Representation
     Redwood Hall  	Pat Hayes (Hayes.pa@xerox.com)
			Abstract below
			
   3:30 p.m.		Tea
     Ventura Hall		


                             --------------
                        THIS WEEK'S CSLI SEMINAR
                      The Acquisition of Morphology
                    Discussion of the debate between
                 Rumelhart/McClelland and Pinker/Prince
                             Dave Rumelhart
                        (der@psych.stanford.edu)
                               October 15

   A couple of years ago Jay McClelland and I developed a connectionist
   model of the process of over-regularization in the acquisition of past
   tense in English.  We were surprised to find how well the model
   accounted for the pattern of such errors children actually make.
   Recently a number of authors who are associated with rather different
   accounts of these results have challenged the adequacy of our model.
   The most massive of these challenges has come from a large paper (150
   pages) by Steve Pinker and Alan Prince.  This paper is to be published
   in the journal COGNITION.  Pinker and Prince challenge our work on
   almost every particular.  Their critique is thoughtful and carefully
   done.  Nevertheless, it seems to me that in their anxiety to dismiss
   our work they missed the point of it.  In my presentation I will: (1)
   sketch our original model (2) sketch the objections raised by Pinker
   and Prince, (3) explain the way in which these objections either miss
   the point of our effort or are simply mistaken and finally (4) offer
   my account of the real significance of our effort and the possibility
   of a connectionist account of linguistic information processing.

                             --------------
                        NEXT WEEK'S CSLI SEMINAR
              External Language and Internal Representation
                                Pat Hayes
                          (Hayes.pa@xerox.com)
                               October 22

   Language evolved, and is used, for communication between intelligent
   agents.  Internally represented information is used quite differently,
   and different assumptions must be made in thinking about ways of
   encoding it for use inside a mind.  In particular, communication can
   assume an intelligent decoder on the other end but is severely
   constrained by the bandwidth of speech, while internal representations
   seem to have much wider channels of communication available between
   their component parts but must be explicit and detailed to an extent
   that would be inappropriate for a `natural' language.  I will argue
   that general talk of `information' ignores this important distinction
   and is therefore sometimes confusing in discussions of situated
   agency.
-------