[Prev][Next][Index]

CSLI Calendar, 19 October 1989, vol. 5:5


       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
-----------------------------------------------------------------------------
19 October 1989                    Stanford                     Vol. 5, No. 5
_____________________________________________________________________________

    A weekly publication of the Center for the Study of Language and
Information (CSLI), Ventura Hall, Stanford University, Stanford, CA 94305-4115
			     ____________

	  CSLI ACTIVITIES FOR THIS THURSDAY, 19 OCTOBER 1989

12:00 noon		TINLunch
      Cordura 100	The Anatomy and Layers of Lexical Meaning
			Julius Moravcsik
			(julius@csli.stanford.edu)
			Abstract in last week's calendar
			
 2:15 p.m.		CSLI Seminar
      Cordura 100	Models of Rational Agency 4
			Michael Bratman, Martha Pollack, Stan Rosenschein
			(bratman@csli.stanford.edu, 
			pollack@warbucks.ai.sri.com, stan@teleos.com)
			Abstract in last week's Calendar
			     ____________

	  CSLI ACTIVITIES FOR NEXT THURSDAY, 26 OCTOBER 1989

12:00 noon		TINLunch
      Cordura 100	Elephant 2000: A Programming Language Based on
			Speech Acts
			John McCarthy
			(jmc@sail.stanford.edu)
			Abstract below

 2:15 p.m.		CSLI Seminar
      Cordura 100	Models of Rational Agency 5
			Michael Bratman, Martha Pollack, Stan Rosenschein
			(bratman@csli.stanford.edu, 
			pollack@warbucks.ai.sri.com, stan@teleos.com)
			Abstract not available
			     ____________

			 NEXT WEEK'S TINLUNCH
      Elephant 2000: A Programming Language Based on Speech Acts
			    John McCarthy

Elephant 2000 is a vehicle for some ideas about programming language
features.

1. Input and output are in an I-O language whose sentences are
   meaningful speech acts approximately in the sense of philosophers
   and linguists.  These include questions, answers, offers,
   acceptances, declinations, requests, permissions, and promises.

2. The correctness of programs is partially defined in terms of
   proper performance of the speech acts.  Answers should be truthful,
   and promises should be kept.  Sentences of logic expressing these
   forms of correctness can be generated automatically from the form
   of the program. 

3. Elephant source programs may not need data structures, because
   they can refer directly to the past.  Thus a program can say that
   an airline passenger has a reservation if he has made one and
   hasn't canceled it.

4. Elephant programs themselves are represented as sentences of
   logic.  Their properties follow from this representation without
   an intervening theory of programming or anything like Hoare
   axioms.

5. Programs that interact nontrivially with the outside world can
   have both _illocutionary_ and _perlocutionary_ specifications,
   i.e., specifications relating inputs and outputs and specifications
   concerning what they do in the world. 
			     ____________

			SYMBOLIC SYSTEMS FORUM
    A Theory of Generalization and a Connectionist Implementation
			    Roger Shepard
		   Friday, 20 October, 3:15, 60:61G
    
After briefly arguing for the central importance of the problem, this
talk will cover three developments.  The first is the empirical
finding of some general and possible universal laws of generalization,
such as (a) that the probability that a response learned in one
situation will be made in a second falls off exponentially with the
distance between the two situations in "psychological" space, and (b)
that the psychological space has a Euclidean or a "city block" metric,
depending on the nature of the relation between its underlying
dimensions.  The second is the formulation of a cognitive theory of
generalization that explains such laws as accommodations to universal
properties of natural kinds in the world in which we have evolved.
The third is the demonstration, through recent connectionist modeling,
that various phenomena of classification learning naturally fall out
of such a cognitive theory.  The conclusion will be suggested that
mental laws, having been shaped by universal regularities of the
world, may partake of some of the universality and elegance of those
regularities.
			     ____________

	    COMMONSENSE AND NONMONOTONIC REASONING SEMINAR
	      A Connotative Treatment of Circumscription
			      Jun Arima
		     ICOT Research Center, Japan
	  Monday, 23 October, 3:15, Margaret Jacks Hall 252

Circumscription proposed by McCarthy is one of the most promising
formalizations of nonmonotonic aspects of commonsense reasoning.  It
has several versions.  However, they are all proposed for the
denotative minimization of predicates, that is, circumscription
minimizes the extension of predicates.  Regarding such treatment, this
paper considers three problems: absence of abnormal things, a
limitation on equality and inconsistency of circumscription.  This
paper proposes a solution for them by presenting a connotative
treatment of circumscription.  This treatment is based on the idea of
circumscribing predicates connotatively, that is, minimizing the set
of names denoting objects that satisfy a certain predicate.  Besides,
this treatment provides a form of the unique names hypothesis for a
general database that allows the use of equality and functions.
			     ____________

		     COMPUTER SCIENCE COLLOQUIUM
	     Algorithms for Matching Patterns in Strings
			    Alfred V. Aho
			AT&T Bell Laboratories
	      Tuesday, 24 October, 4:15, Jordan Hall 040

In the last two decades, striking advances have taken place in the
speed with which we can find patterns in textual data.  This talk will
review the important developments in algorithms for searching for
keywords, sets of keywords, and regular expressions.