Copy Link
Add to Bookmark
Report
NL-KR Digest Volume 01 No. 01
NL-KR Digest (8/11/86 16:23:48) Volume 1 Number 1
Today's Topics:
nl-kr the list
Moderatorial
Recent Advances in KR for Natural Language Processing
Re: On McDermott's "critique of logic"
Request for references
Semantics of Ice Cream
Seminar on Knowledge Bases
Conference - ACL-87
ACL-European Chapter: 1987 CALL FOR PAPERS
----------------------------------------------------------------------
Date: Fri, 1 Aug 1986 19:05 EDT
From: MONTALVO%OZ.AI.MIT.EDU@XX.LCS.MIT.EDU
Subject: nl-kr the list
[Open Letter to the Moderator]
[...]
Could you explain why you are tying together NL and KR, other than
your own personal interests? They are both very large, diverse, and
complex areas of AI. I, in particular, am interested in KR as it
relates and overlaps with vision and graphics. I am not particularly
interested in NL, that is, not in the details of parsing, grammar, and
stuff that is peripheral enough not to overlap with more cetral KR
issues. I am interested in joining the list for the KR content, but I
fear that I will be swamped with messages mostly concerned with NL.
Do you think a person like me should join this list? I'm mostly
interested in visualization, imaging (internal), and symbolic
description of visual knowledge. And am very interrested in KR issues
related to the tie between symbolic descriptions and images. Maybe
the only way to find out is to join and see what kind of list results.
I guess one of the issues is: is this an NL *and* KR list or an NL
*or* KR list? Is it the intersection or union of the two areas?
Fanya Montalvo
------------------------------
Date: Tue, 5 Aug 86 13:07 EDT
From: Brad Miller <nl-kr-request@ROCHESTER.ARPA>
Subject: Moderatorial
Ah, a perfect leading question....
I suppose there are two main reasons that I have tied together NL and KR on
list list, outside of my personal interests.
o Most NL researchers also do KR.
Why? An awful lot of current NL work is in fact beyond the parsing
and grammar stage. (This is not to disparage the efforts of those who continue
to do work in this area, since such topics are indeed part of the focus of
this list). Most "high-level" NL deals with pragmatics, contexts, beliefs,
speech acts, each of which requires new approaches to traditional KR problems.
As McDermott's paper "A Critique of Pure Reason" points out, logic may not be
enough. And this is not just the logic of the old "big crank" theory, but more
modern attempts to extend it as well. Whatever turns out to be the "best" KR,
it is clear that NL is about the world, which implies that agents must somehow
represent that world for themselves. The implication is, then, that solutions
to NL require appropriate solutions to KR.
o This list is a "sublist" of AIList.
What I mean by this, is that given the entire range of topics AIList
covers, I didn't want to split off just one small subarea. That would neither
help the large amount of traffic on AIList, nor secure the success of the
smaller list. NL and KR seem interrelated enough (to me, at least) that the
clique it proscribes may be useful in some sense -- that is, the communication
between issues inside and outside the list are more minimal than had I chosen
a smaller subject area.
As for KR being done in other areas, terrific! We want to hear about
it, unless there is a better list for it (if you are doing low-level vision
representations that could not be carried over to much of anything else, I
would suggest using the vision list, for example). Why would you want to be on
this combined list if you are ONLY interested in KR? Well, I think that many
people doing high-level vision, and other topics, realize that a lot of the
problems they are trying to solve are the same ones the NL people already have
come up against in their work. In my opinion, high level vision is a LOT like
NL - the mapping of "parsing" scenes, assigning "semantic values" to
conglomerations of features (object detection) and then looking at the whole
(pragmatics) seems reasonable and productive.
Certainly there will be some things posted that you are not
particularly interested in. But that would be true, even if the list were
devoted exclusively to KR! I intend to arrange the digests in a manner that
will allow one to skim or skip discussions and topics that are not of central
interest to the reader. I hope, of course, that what will be interesting and
of value will justify the existance of the list and everyone's contributions
to it. But lists have some degree of inertia - the list can be no better (and
no worse) than the contributions to it. All I can do as editor/moderator is
attempt to weed out topics and discussions that are obviously irrelevant or
trivial, and format the remainder into a digestable whole. I cannot "referee"
every submission for good technical content or high standards, because I don't
have the time. The list itself, however, can police itself if what it wants is
these high standards - by submitting only carefully thought out articles, and
violently objecting to submissions that are substandard. I don't think this is
necessarily a good idea, since part of the task of the list as I see it is to
teach - to allow us to put up some half baked schemes as a target, and then
learn from their weaknessess. Otherwise we will just be duplicating ACL or
other fine journals - and have corresponding delays in the reporting of
results, and relative infrequency of publication.
In short, my hope is that this will be a wider communications channel
than AIList, because of it's narrower scope. I WANT to hear what the language
philosophers have to say about our KR ideas. I HOPE that grad students just
starting out in the area will use the list to ask questions, and discuss their
thoughts, perhaps in preparation for their area exams or dissertations. And
that those amoung us who are more senior will help guide the discussions to
fruitful conclusions, or at least a better understanding of the problems we
look at. That the hackers amoung us will get some good ideas for implementing
solutions. And so on. But ultimately, this is YOUR list. You all will have to
make it what you want it to be.
Brad Miller
nl-kr-request@rochester.arpa
------------------------------
Date: 9 Aug 1986 17:18:48 PDT
From: Norm Sondheimer <Sondheimer@B.ISI.EDU>
Subject: Recent Advances in Knowledge Representation for Natural Language Processing
Ralph Weischedel and I are giving a tutorial on Recent Advances in Knowledge
Representation for Natural Language Processing at Coling this month. It
struck us that our opinions on what was significant may not be shared by
everyone. If anyone cares to venture an opinion, we will be glad to respond
with our outline.
Incidentally, it also struck us that not everyone would agree on what
constitutes a knowledge representation. We are taking formal logic as the
prototypical representation and branching out from there.
Norm Sondheimer
Sondheimer@b.isi.edu
------------------------------
Date: 4 Aug 86 14:00:26 GMT
From: jay@rochester.ARPA (Jay Weber)
Subject: Re: On McDermott's "critique of logic"
[Forwarded from ROCHESTER's BBoard by miller@ur-acorn.arpa]
>From: Jun Tarui <tarui>
> Everyone (including McDermott) seems to
>agree that first-order logic has a well-defined semantics i.e. model theory;
>predicates being mapped to relations, etc. So essentially you are giving
>"meaning" in terms of set theory. But what if we ask further what these
>sets and elements *mean*? You may say we can't keep on asking forever.
>Well, then, what exactly is it that allows us to claim that first-order logic
>has a clear semantics? In other words, how do you *define* (or is it possible
>to *define*) the system S having a semantics?
Unfortunately, we cannot relate any formalism to the real world, only
other formalisms. The problem is establishing reference; even when we
say "X stands for that chair" we have really only related "X" and
"that chair", and counted on the natural connection between "that chair"
and that chair. Brian Smith in his paper "The Correspondence Continuum"
criticizes the common belief that model theory is the be-all end-all of
semantics. The paper is certainly causes you to ponder what it means
to provide a semantics, but he doesn't provide a viable alternative to
model theory. Anyway, the assumption with model theory is that sets
and set membership naturally correspond with aspects of the real world.
Jay
------------------------------
Date: Fri, 8 Aug 86 11:33 EDT
From: Gray Abbott <harvard!dartvax!creare!gda>
Subject: Request for references
We have a seminar group which meets weekly to discuss the
literature in the AI field. Lately we've been reading a lot
of papers on Expert Systems. Many members of the group have
complained that the literature is rather vague when discussing
Knowledge Acquisition; we would like more detail on what
knowlege engineers and experts say to each other. So far we
have been unable to find any references which satisfy our
wants. Can anyone recommend any books or papers (try to
include the name and date of the journal) that address this?
Gray Abbott
Creare Inc.
Hanover, NH
{...dartvax!creare!gda}
------------------------------
From: George Cross <FACCROSS%WSUVM1.BITNET@WISCVM.ARPA>
Subject: Semantics of Ice Cream
Take a look at the cover of the August 4, 1986 issue of the New Yorker
magazine. An erudite scientist is shown lecturing about a taxonomy of
ice cream types including sandwiches, cones, etc. I think this could be
an effective cover for a KR book -- perhaps the second edition of "Readings
in KR" by Brachman and Levesque. The cartoon is by Roz Chast.
---- George
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
George R. Cross cross@wsu.CSNET
Computer Science Department cross%wsu@csnet-relay.ARPA
Washington State University faccross@wsuvm1.BITNET
Pullman, WA 99164-1210 (509)-335-6319/6636
------------------------------
Date: Fri, 8 Aug 86 17:25 EDT
From: Brad Miller <nl-kr-request@ROCHESTER.ARPA>
Subject: Seminar on Knowledge Bases
[Forwarded from IRList by miller@ur-acorn.arpa]
Date: Friday, 18 July 1986 10:09-EDT
from: BGOODMAN@G.BBN.COM
Subject: Seminar on Knowledge Bases
BBN Laboratories Inc.
Science Development Program
AI/Education Seminar
Friday, 1 August 10:30am
From Guidon to Neomycin and Heracles--Viewing
Knowledge Bases as Qualitative Models
Dr. William J. Clancey
Stanford Knowledge Systems Laboratory
Computer Science Department
701 Welch Road, Bldg C
Palo Alto, CA 94304
Beginning with early attempts to improve MYCIN's representation of
knowledge for use in teaching, we have followed the approach of
decomposing knowledge from how it is used, abstracting knowledge
structures and reasoning procedures, and formulating an increasingly more
general understanding of what knowledge engineering and knowledge bases
are all about. In NEOMYCIN, medical knowledge and diagnostic procedure
are separately represented in well-structured languages to facilitate
explanation and student modeling. In HERACLES, this knowledge base is
viewed as a classification model of some physical, cognitive, or social
system that is heuristically related to some design, modification,
prediction, or control action. That is, we view the knowledge base as a
qualitative model of some system in the world, designed with practical
engineering value in mind. From this perspective, the "diagnostic
strategy" of Neomycin is a general inference procedure that describes
memory activation and search for constructing a situation-specific model.
This talk will review the development of NEOMYCIN from GUIDON and
summarize the generalizations that we are now exploiting in our
development of the HERACLES shell and GUIDON2 teaching programs.
------------------------------
Date: Sat, 9 Aug 86 15:36:23 edt
From: walker@mouton.bellcore.com (Don Walker)
Subject: Conference - ACL-87
CALL FOR PAPERS
25th Annual Meeting of the Association for Computational Linguistics
6-9 July 1987
Stanford University
Stanford, California, USA
TOPICS OF INTEREST: Papers are invited on substantial, original, and
unpublished research on all aspects of computational linguistics,
theoretical or applied. Papers that demonstrate the power of a
technique for computational purposes or present new ideas together with
a comparison and evaluation of their implications for current research
are particularly welcomed. The following topics are illustrative:
phonology, morphology, the lexicon, syntax, semantics, discourse, pragmatics;
parsing and generation; language understanding; knowledge representation;
speech acts and planning; language interfaces; language acquisition;
speech analysis and synthesis; information and document retrieval;
computational, mathematical, and psychological models;
programming strategies; computer architectures.
REQUIREMENTS: Papers should describe unique work that has not been
submitted elsewhere; they should emphasize completed work rather than
intended work; and they should indicate clearly the state of completion
of the reported results.
FORMAT FOR SUBMISSION: Authors should submit ten copies of an extended
abstract not to exceed eight double-spaced pages (exclusive of
references) in a font no smaller than 10 point (elite). The title page
should include the title, the name(s) of the author(s), complete
addresses, a short (5 line) summary, and a specification of the topic
area. Submissions that do not conform to this format will not be
reviewed. Send to:
Candy Sidner, ACL-87 Program Chair
BBN Laboratories Inc.
10 Moulton Street
Cambridge, MA 02238, USA
(617)497-3566; sidner@g.bbn.com
SCHEDULE: Papers must be received by 12 January 1987.
Authors will be notified of acceptance by 2 March 1987.
Camera-ready copies of final papers prepared in a double-column
model paper format must be received by 30 April, along with a
signed copyright release statement.
OTHER ACTIVITIES: Special events will signal the 25th Anniversary of
the ACL. The meeting will also include a program of applied tutorials
and a variety of exhibits and demonstrations. Anyone wishing to
arrange an exhibit or present a demonstration should send a brief
description together with a specification of physical requirements
(space, power, telephone connections, tables, etc.) to Paul Martin,
SRI International, Menlo Park, CA 94025, USA; (415)859-4480;
pmartin@sri-ai.arpa.
CONFERENCE INFORMATION: Local arrangements are being handled by
Doug Appelt, SRI International, Menlo Park, CA 94025; (415)859-6150;
appelt@sri-ai.arpa. For other information on the conference and on the
ACL more generally, contact Don Walker (ACL), Bell Communications
Research, 445 South Street, MRE 2A379, Morristown, NJ 07960;
201:829-4312; walker@mouton.bellcore.com or walker%mouton@relay.cs.net
or bellcore!walker@ucbvax.berkeley.edu.
LSA SUMMER LINGUISTIC INSTITUTE: ACL-87 will be preceded and followed
by the 54th LSA Institute at Stanford University, which will be
cosponsored by the ACL and the AAAI. Computational linguistics will be
the major focus for the Institute. The week preceding the 25th Annual
Meeting will feature ten 8-hour courses that can be taken by ACL
members as tutorials. For more information, contact Ivan Sag,
Department of Linguistics, Stanford University, Stanford, CA, 94305,
USA; (415)723-4284; institute@su-csli.stanford.edu. A conference on
Logic and Linguistics will follow ACL-87 from 10-11 July, also at
Stanford University; contact Rich Thomason, Linguistics Department,
University of Pittsburgh, Pittsburgh, PA 15260; (412)624-5791;
thomason@c.cs.cmu.edu.
------------------------------
Date: Sat, 9 Aug 86 15:52:36 edt
From: walker@mouton.bellcore.com (Don Walker)
Subject: ACL-European Chapter: 1987 CALL FOR PAPERS
CALL FOR PAPERS
Third Conference of the European Chapter
of the Association for Computational Linguistics
1-3 April 1987
Copenhagen University
Copenhagen, Denmark
PURPOSE: This conference is the third in a series of biennial
conferences on computational linguistics sponsored by the European
Chapter of the Association for Computational Linguistics. It was
preceded by meetings in Pisa in September 1983 and in Geneva in March
1985. Although regional in organizational structure, these conferences
are global in scope and participation.
SCOPE: Papers are invited on all aspects of computational linguistics,
including, but not limited to:
morphology;
lexical semantics;
computational models for the analysis and generation of language,
spoken and written;
computational lexicography and lexicology;
syntax and semantics;
machine translation and translation aids;
natural language interfaces, knowledge representation, and expert systems.
SUBMISSION: Authors should send seven copies of a 5 to 8 page double-
spaced summary to the Programme Committee Chairperson:
Bente Maegaard
Institut for Anvendt og Matematisk Lingvistik
Kobenhavns Universitet
Njalsgade 96
DK-2300 Kobenhavn S, DENMARK
Telephone: +45-1-542 211
It is important that the summary identify the new ideas in the paper
and clearly indicate to what extent the work is complete and to what
extent it has been implemented. It should contain sufficient
information to allow the programme committee to determine the scope of
the work and its relation to relevant literature.
SCHEDULE: Summaries must be submitted by 15 November 1986. Authors
will be notified of acceptance by 1 February 1987. Camera-ready copies
of final papers prepared in a double-column model paper format must be
received by 3 April 1987, that is, at the end of the Conference, along
with a signed copyright release statement. The Proceedings will be
published after the Conference.
EXHIBITS AND DEMONSTRATIONS: A program of exhibits and demonstrations
is planned. Anyone wishing to participate is asked to contact Bente
Maegaard.
------------------------------
End of NL-KR Digest
*******************
