[Prev][Next][Index]

Newsletter September 5, No. 44



                      C S L I   N E W S L E T T E R
_____________________________________________________________________________
September 5, 1985               Stanford                       Vol. 2, No. 44
_____________________________________________________________________________
                                
     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, September 5, 1985

   12 noon		TINLunch
     Ventura Hall       ``Predication, Logical Syntax, and the Type-free 
     Conference Room    Conception of Properties, Relations, and Propositions''
			Discussion led by Chris Menzel

   2:15 p.m.		CSLI Talk
     Ventura Hall	``FORK: A Flavor-Based Environment for Object-oriented
     Seminar Room	Knowledge Representation''
			C. Beckstein, G. Goerz, 
			University Erlangen-Nuernberg, W. Germany 
			(Abstract on page 2)

   3:30 p.m.		Tea
     Ventura Hall		

                              ____________

         CSLI ACTIVITIES FOR *NEXT* THURSDAY, September 12, 1985

   12 noon		TINLunch
     Ventura Hall       ``Free Word Order in GPSG'' 
     Conference Room    by Arnold Zwicky
			Discussion led by Hans Uszkoreit, SRI and CSLI
			(Abstract on page 2)

   2:15 p.m.		CSLI Talk
     Ventura Hall	``Arithmetical Truth and Hidden Higher-Order Concepts''
     Seminar Room	Daniel Isaacson, Oxford University
			(Abstract on page 2)

   3:30 p.m.		Tea
     Ventura Hall		


Page 2                     CSLI Newsletter                  September 5, 1985
_____________________________________________________________________________
                   ABSTRACT FOR THIS WEEK'S CSLI TALK
                 ``FORK: A Flavor-Based Environment for
               Object-oriented Knowledge Representation''
   C. Beckstein, G. Goerz, University Erlangen-Nuernberg, West Germany
            2:15, Thursday, September 5, Ventura Seminar Room

      Most object-oriented extensions of LISP provide only marginal
   support for the purpose of knowledge representation. In particular,
   there are only poor means---if any---for specifying meta-information
   about attributes of objects such as typed domains, methods for
   determining values (demons), multiple-valued attributes and explicit
   control of inheritance.  Furthermore, they usually don't offer
   adequate utilities for handling multiple perspectives, retrieving
   objects through patterns of characteristic features, and maintaining
   structural relations (integrity constraints) in and between objects.
   FORK is an attempt to extend Flavors, an object-oriented extension of
   LISP, by adding features which are well known from frame-like systems
   with the advantage of keeping a systematic distinction between classes
   and instances. The procedural knowledge is attached to classes either
   in the usual sense of methods as functions or in the form of (forward
   chaining) rule sets. In addition, FORK offers a programming
   environment to support users in the construction and maintenance of
   large, hybrid knowledge bases.
                              ____________
                    ABSTRACT FOR NEXT WEEK'S TINLUNCH
                       ``Free Word Order in GPSG''

      The ID/LP version of GPSG provides an elegant scheme for describing
   certain syntactic phenomena that are usually subsumed under the terms
   ``free word order'' or ``free constituent order.''  The questions
   addressed in this paper are (i) whether (and how) all the variants and
   degrees of ordering freedom can be described in the framework and (ii)
   whether universal generalizations can be expressed.  In this connection, 
   a universal version of a Pullum-type liberation rule is discussed.
                              ____________
                   ABSTRACT FOR NEXT WEEK'S CSLI TALK
         ``Arithmetical Truth and Hidden Higher-Order Concepts''
                   Daniel Isaacson, Oxford University
          2:15, Thursday, September 12, Ventura Conference Room

      The natural numbers can be characterized within a given domain by
   the second-order condition that they constitute a minimal collection
   closed under a given one to one mapping (succession) and containing an
   element (zero) not in the range of that mapping.  Peano Arithmetic is
   what can be expressed of this characterization by first-order
   axiomatization, and is in this sense a natural, conceptually intrinsic
   formal system.  On the other hand, Godel showed that the truths of
   arithmetic are not recursively enumerable, so that any true axiomatic
   formal system for arithmetic has proper first-order extensions.  It
   might seem in this way that no one first-order axiomatic system could
   be of intrinsic conceptual importance.  This talk will explore
   considerations that might resolve the tension between these two
   observations, in particular to suggest that possibly Peano Arithmetic
   can be considered as complete with respect to genuinely arithmetical
   truth, in the sense that perceiving the truth of first-order truths in
   the language of arithmetic that are not provable in Peano Arithmetic
   must be in terms of higher-order concepts which they code.

Page 3                     CSLI Newsletter                  September 5, 1985
_____________________________________________________________________________
                                  TALK
            ``Unification-Based Speech Parsing with a Chart''
     C. Beckstein, G. Goerz, Univ. Erlangen-Nuernberg, West Germany
            2:15, Tuesday, September 10, Ventura Seminar Room

      We describe GuLP, a chart parser to be used as a syntactic module
   of the Erlangen Speech Understanding System EVAR. Operating with a
   unification grammar, GuLP realizes an agenda-based multiprocessing
   scheme, which allows the application of various parsing strategies to
   fragments of the same utterance in a transparent way.  The overall
   control mechanism is realized through a general interrupt system.  In
   order to process speech data, a variety of new features has been
   incorporated: in particular the ability to perform incremental
   analysis, to do direction independent island parsing, to process gaps
   in utterances, and to handle hypothesis scores. Finally, a complexity
   estimate and a few experimental results are discussed.
                              ____________
            INTERACTIONS OF MORPHOLOGY, SYNTAX, AND DISCOURSE
                    ``Discourse-based Interactions of
           Four Morphosyntactic Subsystems in Northern Pomo''
              Summary of the meeting on Thursday, August 29

      Formal theories of discourse representation face the task of
   modelling both text-external deixis (expressions anchored to the
   context of utterance) and text-internal deixis (e.g., Partee's (1984)
   treatment of Reichenbach's `reference time' or Banfield's (1982)
   discourse primitive SELF.)  Cathy O'Connor discussed these dimensions
   in the light of complex interactions of four morphosyntactic
   subsystems of Northern Pomo.
   (1) Third person, non-clause-bounded reflexives function
       logophorically; they establish that the SELF is anchored
       to 3rd person.
   (2) A possessive prefix found on kinship terms that is
       necessarily a bound anaphor is optional outside the
       minimal clause containing its antecedent.  If SELF is 3rd
       person, the anaphoric prefix is obligatory, deictically
       linked to this text-internal discourse entity.
   (3) An alternation in case-marking for subjects of
       unaccusative verbs conveys subjective expression of
       internal experience (`I'm feeling really sick') versus
       objective reporting (`I'm sick today'). This is normally
       limited to 1st person subjects. In discourse contexts
       where SELF is 3rd person, the alternation is sanctioned
       for 3rd person subjects.
   (4) A set of `evidential' verbal inflections, which
       indicate utterance-speaker's evidence for the assertion,
       display pragmatically motivated co-occurrence restrictions
       with respect to the above phenomena.
      In light of these findings the relation of logophoricity and
   subjunctive mood was discussed. The proposal was made that mood
   subordination in a discourse representation is the appropriate domain
   for an account of these complex facts.  Finally, the problem of
   representing (1) through (4) above was discussed, and the notions of
   text-internal and text-external deixis were suggested to be necessary
   components in a representation of the discourse structures
   involved.					--Cathy O'Connor
-------