[Prev][Next][Index]

Newsletter, Dec. 6, No. 7



                       C S L I   N E W S L E T T E R
___________________________________________________________________________
December 6, 1984               Stanford                      Vol. 2, No. 7
___________________________________________________________________________

         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, DECEMBER 6, 1984

12 noon                 TINLunch
  Ventura Hall          Fodor's ``Psychosemantics''
  Conference Room       Discussion led by Ned Block.

2:15 p.m.               CSLI Seminar
  Redwood Hall          ``The Structures of Discourse Structure''
  Room G-19             by Barbara J. Grosz
                        Discussant will be Ray Perrault.

3:30 p.m.               Tea
  Ventura Hall		

4:15 p.m.               CSLI Colloquium
  Redwood Hall          Discussion of DOD Funding
  Room G-19             Donald Kennedy, President, Stanford
                        Sydney Drell, Dep. Dir., SLAC
                        John Etchemendy, discussion leader
                           ____________ 

                SCHEDULE FOR ***THURSDAY, DECEMBER 13,*** 1984

12 noon                 TINLunch
  Ventura Hall          ``Syntactic Features, Semantic Filtering,
  Conference Room       and Generative Power''
                        Discussion led by Peter Sells

2:15 p.m.               CSLI Seminar
  Redwood Hall          ``A Generalized Framework for Speech Recognition''
  Room G-19             by Marcia Bush
                        Discussant will be Ray Perrault.
                        (Abstract on page 2)

3:30 p.m.               Tea
  Ventura Hall		

4:15 p.m.               CSLI Colloquium
  Redwood Hall          ``Data Semantics''
  Room G-19             Fred Landman, Department of Philosophy,
                          University of Amsterdam
                        Peter Sells, discussion leader
                        (Abstract on page 2)
Page 2                   CSLI Newsletter                   December 6, 1984
___________________________________________________________________________

                ABSTRACT OF NEXT WEEK'S SEMINAR
      ``A Generalized Framework for Speech Recognition''

This talk will describe a framework for speaker-independent,
large-vocabulary and/or continuous speech recognition being developed at
Schlumberger (Fairchild).  The framework consists of three components:
  1) a finite-state pronunciation network which models relevant
     acoustic-phonetic events in the recognition vocabulary;
  2) a set of generalized acoustic pattern matchers; and
  3) an optimal search strategy based on a dynamic programming algorithm.
The framework is designed to accommodate a variety of (typically disparate)
approaches to the speech recognition problem, including spectral template
matching, acoustic-phonetic feature extraction and lexical pruning based
on broad-category segmentation.  A working system developed within this
framework and tailored to the digits vocabulary will also be described.  The
system achieves high recognition accuracy on a corpus spoken by
approximately 250 talkers from 22 ``dialect groups'' within the continental
United States.
                                                ---Marcia Bush
                        ____________

                ABSTRACT OF NEXT WEEK'S COLLOQUIUM
                      ``Data Semantics''

Abstract: There is a growing agreement of opinion that several semantic
phenomena can only be adequately dealt with in a theory which takes
partiality seriously, a theory of partial objects. There is no agreement
about what these partial objects are; for instance, whether they represent
``pieces of the world'' or ``states of partial information about the world.''
Yet, the choice of the perspective determines in large part the potential
of the theory.  I will discuss various aspects of Data Semantics, a theory
being developed by Frank Veltman and me, which takes the second
perspective as basic: the semantic behavior of several types of expressions
can best be understood if we take them to relate to our lack of information,
and regard them as patterns on how information can grow. I will argue that
problems concerning quantification and equality force us to distinguish
between different kinds of partial objects.
                                                        ---Fred Landman
                        ____________

                SUMMARY OF LAST WEEK'S SEMINAR
                 ``Parsing Acoustic Events''
                      Meg Withgott, CSLI

Withgott described a system for parsing low-level acoustic events using
language-independent input and language-specific formal grammars.  This
decomposition of the identification/recognition problem, she argued, permits
a representation of physical signals that is stable over utterance
situations and a way to extract information from them using familiar
techniques from higher-level natural language research.  Alex Pentland, who
served as the commentator, remarked on some similarities between this view
of spoken language and current work on vision.

Page 3                   CSLI Newsletter                  December 6, 1984
___________________________________________________________________________

                    F1 (AND F3) PROJECT MEETING

Title:     Self-propagating Search of Memory
Speaker:   Pentti Kanerva
Time/Date: Tuesday, December 11, 3:15 p.m.
Place:     Ventura Seminar Room

Abstract: Human memory has been compared to a film library that is indexed
by the contents of the film strips stored in it.  How might one construct
a computer memory that would allow the computer (a robot) to recognize
patterns and to recall sequences the way humans do?  The model presented
is a simple generalization of the conventional random-access memory of a
computer.  However, it differs from it in that (1) the address space is very
large (e.g., 1,000-bit addresses), (2) only a small number of physical
locations are needed to realize the memory, (3) a pattern is stored by
adding it into a SET of locations, and (4) a pattern is retrieved by POOLING
the contents of a set of locations.  Patterns (e.g., of 1,000 bits) are
stored in the memory (the memory locations are 1,000 bits wide) and they
are also used to address the memory.  From such a memory it is possible to
retrieve previously stored patterns by approximate retrieval cues--thus,
the memory is sensitive to similarities.  By storing a sequence of patterns
as a linked list, it is possible to index into any part of any "film strip"
and to follow the strip from that point on (recalling a sequence).
                         ____________

                       AREA C MEETING

Topic:     Theories of variable types for mathematical practice,
           with computational interpretations
Speaker:   Solomon Feferman, Depts. of Mathematics and Philosophy
Time/Date: 1:30-3:30 p.m., Wednesday, December 12
Place:     Conference Room, Ventura Hall

Abstract:  A new class of formal systems is set up with the following
characteristics:
   1) Significant portions of current mathematical practice (such as in
      algebra and analysis) can be formalized naturally within them.
   2) The systems have standard set-theoretical interpretations.
   3) They also have direct computational interpretations, in which all
      functions are partial recursive.
   4) The proof-theoretical strengths of these systems are surprisingly
      weak (e.g. one is of strength Peano arithmetic).
   Roughly speaking, these are axiomatic theories of partial functions and
classes.  The latter serve as types for elements and functions, but they
may be variable (or "abstract") as well as constant.  In addition, an element
may fall under many types ("polymorphism").  Nevertheless, a form of typed
lambda calculus can be set up to define functions.
   The result 3) gets around some of the problems that have been met with
the interpretation of the polymorphic lambda calculus in recent literature
on abstract data types.  Its proof requires a new generalization of the
First Recursion Theorem, which may have independent interest.
   The result 4) is of philosophical interest, since it undermines 
arguments for impredicative principles on the grounds of necessity for
mathematics (and, in turn, for physics).
   There are simple extensions of these theories, not meeting condition 2),
in which there is a type of all types, so that operations on types appear
simply as special kinds of functions.
Page 4                   CSLI Newsletter                   December 6, 1984
___________________________________________________________________________

                           NL1 MEETING

Topic:      ``Association with Focus''
Speaker:    Mats Rooth
Time/Date:  2 p.m., Friday, December 7
Place:      Trailer Seminar Room
Note:       The content will overlap with but be non-identical to the
            presentation the speaker gave in the intonation seminar.

Abstract: In the context of adverbs of quantification, conditionals, and
``only,'' focus can have truth conditional significance.  Suppose Mary
introduced Bill and Tom to Sue and performed no other introductions.  Then
``Mary only introduced Bill to SUE'' is true, while ``Mary only introduced
BILL to Sue'' is false.  Similarly, ``MARY always takes Sue to the movies''
and ``Mary always takes SUE to the movies'' have different truth conditions.
My general claim is that focus influences truth conditions indirectly:  the
semantics of the constructions in question involve contextual parameters,
typically unspecified domains of quantification, which are fixed by a
focus-influenced component of meaning.  This idea is executed in a Montague
grammar framework.
                           ____________ 

        *****THURSDAY CSLI ACTIVITIES BETWEEN QUARTERS*****
There will be no center-wide Thursday activites between quarters (on
December 20, 27, and January 3). Regular Thursday activities will resume
on January 10.

-------