[Prev][Next][Index]

CSLI Calendar, January 28, 3:15


       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
_____________________________________________________________________________
28 January 1988                    Stanford                    Vol. 3, No. 15
_____________________________________________________________________________

     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, 28 January 1988

   12 noon		TINLunch
     Ventura Hall       Reading: "True Believers: The Intentional
     Conference Room  	Strategy and Why it Works"  
			by Daniel Dennett
			Discussion led by Adrian Cussins
			(cussins.pa@xerox.com)
			Abstract in last week's Calendar

   2:15 p.m.		CSLI Seminar
     Room G-19          Modal Subordination, Situations, and Reference Time
     Redwood Hall  	Craige Roberts
			(croberts@csli.stanford.edu)
			Abstract in last week's Calendar
			
   3:30 p.m.		Tea
     Ventura Hall		
                             --------------
	   CSLI ACTIVITIES FOR NEXT THURSDAY, 4 February 1988

   12 noon		TINLunch
     Ventura Hall       Reading: "Some Uses of Higher-Order Logic
     Conference Room  	in Computational Linguistics"
			by Dale A. Miller and Gopalan Nadathur
			Discussion led by Douglas Edwards
			(edwards@warbucks.ai.sri.com)
			Abstract below

   2:15 p.m.		CSLI Seminar
     Room G-19          A Nonmonotonic Account of Causation
     Redwood Hall  	Yoav Shoham
			(shoham@score.stanford.edu)
			Abstract below
			
   3:30 p.m.		Tea
     Ventura Hall		
                             --------------
			  NEXT WEEK'S TINLUNCH
		Reading: "Some Uses of Higher-Order Logic
		      in Computational Linguistics"
		 by Dale A. Miller and Gopalan Nadathur
     from "24th Annual Meeting of the Association for Computational
	   Linguistics: Proceedings of the Conference" (1986)
		    Discussion led by Douglas Edwards
		      (edwards@warbucks.ai.sri.com)
			     4 February 1988

   Miller and Nadathur present a system of higher-order logic (typed
   lambda-calculus) as a suitable formalism for the representation of
   syntactic and semantic information in computational linguistics.  They
   argue that such a formalism is clearer and more natural than available
   alternatives.  They also reply point by point to certain standard
   criticisms of the computational use of higher-order logic.  In
   particular, they argue that:

   (1) Theoretical linguistics is often heavily committed to higher-order
       logic anyway (Montague Semantics, for example) and it will be
       easier to design working systems to fit a theory if the
       computational formalism mirrors the ontology of the theory.

   (2) Even if a first-order formalism is used to represent the semantics
       of sentences, *reasoning* about semantics is an inherently
       higher-order process and cannot be represented with full
       naturalness in the same formalism.  This fact leads to the use of
       ad hoc procedures for semantics and to the development of separate
       semantic and syntactic formalisms.  The use of higher-order logic
       allows easier integration of semantic and syntactic processing.

   (3) The formalization of semantic processing in first-order formalisms
       like Prolog is bedevilled by the need to consider explicitly the
       intricate processes of substitution and variable binding.  A logic
       programming language for higher-order logic, like Miller and
       Nadathur's LambdaProlog, obviates this need through the use of
       beta-conversion in the language itself.

   (4) The difficulty of theorem-proving in higher-order logic is evaded
       by confining attention to a restricted set of formulas (analogous
       to Horn clauses in first-order logic) and lowering sights from
       full theorem-proving to logic programming, using a highly
       restricted proof procedure.  If more is needed, restricted
       theorem-provers can easily be designed *within* LambdaProlog.

   It would also appear that much ordinary reasoning even outside of
   linguistic semantics is higher-order.  Are Miller and Nadathur right
   in thinking that their formalism can help to bridge the gap between
   linguistic theory and computational practice?
			     --------------
			   NEXT WEEK'S SEMINAR
		   A Nonmonotonic Account of Causation
			       Yoav Shoham
		       (shoham@score.stanford.edu)
			     4 February 1988

   We suggest that taking into account considerations that traditionally
   fall within the scope of computer science in general, and artificial
   intelligence in particular, may shed light on the concept of
   causation. We argue that causal reasoning is a mechanism for making
   coarse, but fairly reliable, inferences in the absence of full
   information. Specifically, we propose that the concept of causation is
   intimately bound to that of nonmonotonic reasoning. We offer an
   account of causation which relies on this connection, and briefly
   compare our proposal to previous accounts of causation within
   philosophy.