Copy Link
Add to Bookmark
Report
NL-KR Digest Volume 13 No. 50
NL-KR Digest Mon Nov 14 23:35:14 PST 1994 Volume 13 No. 50
Today's Topics:
CFP: ASIS SIG/ALP (NLP) Automated Language Processing, Oct 95, Chicago
CFP: 5th Toulouse Wkshp Time, Space and Movement, Jun 95, Gascony
CFP: PACLING '95, Pacific Rim NLP Wkshps, Apr 95, Brisbane
Announcement: Wkshp on context (NLP, KR, etc.), Aug 95, Montreal
Announcement: Report on Connectionism, Semantics
* * *
Subcriptions: listserv-style administrative requests to
nl-kr-request@ai.sunnyside.com.
Submissions, policy, questions: nl-kr@ai.sunnyside.com
Back issues:
FTP: ai.sunnyside.com:/pub/nl-kr/Vxx/Nyyy
/pub/nl-kr/Vxx/INDEX
Gopher: ai.sunnyside.com, Port 70, in directory /pub/nl-kr
Email: write to LISTSERV@AI.SUNNYSIDE.COM, omit subject, mail command:
GET nl-kr nl-kr_file_list
Web: http://ai.sunnyside.com/pub/nl-kr
Editors:
Al Whaley (al@ai.sunnyside.com) and
Chris Welty (weltyc@cs.vassar.edu).
-----------------------------------------------------------------------
Date: Tue, 08 Nov 94 08:50:14 EST
From: "Coble, Jim" <jim@mail.lib.duke.edu>
To: nl-kr@ai.sunnyside.com
Subject: CFP: ASIS SIG/ALP (NLP) Automated Language Processing, Oct 95, Chicago
CALL FOR PAPERS
The Automated Language Processing Special Interest Group (SIG/ALP) of
the American Society for Information Science (ASIS) invites submission
of papers on the topics listed below for panel sessions at the 1995
ASIS Annual Meeting in Chicago, IL, October 9-12. Papers can present
theoretical perspectives, experimental results, descriptions of
completed research or research in progress, critical evaluations, or
demonstrations of pertinent products. Please submit a description of
your topic (200 words or less) by November 30, 1994, to
= = = = = = = = = = = = = =
Jim Coble Voice: 919-660-5859
Senior Systems Librarian Fax: 919-684-2855
Perkins Library Internet: jim@mail.lib.duke.edu
Box 90196
Duke University
Durham, NC 27708-0196
= = = = = = = = = = = = = =
Please include your name, job title, company, address, telephone
number, fax number, and e-mail address. Notification of acceptance
will be sent by February 15, 1995. If you have any questions, please
also feel free to contact me as shown above. We would also welcome
inquiries and proposals from other ASIS SIG's that may want to
contribute perspectives or jointly sponsor sessions on these topics.
* Natural Language Tools for Navigating the Internet
SIG/ALP seeks presentations on natural language tools for retrieving
information on the Internet. Of particular interest are presentations
that include an evaluation of the strengths and weaknesses of these
tools for navigating the Internet.
* Use of Natural Language Processing in Multimedia Systems
SIG/ALP seeks presentations on the actual and potential use of natural
language processing capabilities in multimedia systems. Possible
topics include the use of accompanying natural language text to
facilitate retrieval of non-textual data, challenges and opportunities
in automated indexing of multimedia databases, and the automated
generation of explanatory information in mutimedia systems.
* Instructional Uses of Automated Language Processing
SIG/ALP seeks presentations on the actual and potential use of natural
language understanding in computer-assisted instruction, intelligent
tutoring systems, and other interactive educational systems.
* Current Research in Natural Language Processing
SIG/ALP invites persons to share the results of completed research or
research in progress in the field of automated language processing at
a panel session intended to highlight topics of current interest to
natural language processing researchers.
-----------------------------------------------------------------------
To: aisb@cogs.sussex.ac.edu, bulletin-PRC-IA@irisa.fr,
Subject: CFP: 5th Toulouse Wkshp Time, Space and Movement, Jun 95, Gascony
Date: Wed, 09 Nov 1994 18:12:16 +0000
From: Pascal AMSILI <amsili@gogol.irit.fr>
5th Toulouse International Workshop
TIME, SPACE and MOVEMENT
-----
Meaning and Knowledge in the Sensible World
Organized by the ``Langue, Raisonnement, Calcul'' Group
IRIT, Universite Paul Sabatier Toulouse
ERSS, Universite de Toulouse-Le Mirail
CNRS URA 1399, URA 1033
Gascony (near Toulouse), France
23-28 June, 1995
~~~~~~~~~~~~~~~
CALL FOR PAPERS
~~~~~~~~~~~~~~~
This workshop will be the fifth one in a series which began in 1989.
Like previous editions, it aims at gathering researchers from a
variety of fields around the themes of the semantics of Time, Space
and Movement, in a castle in the middle of the beautiful landscape of
Gascony. Unlike previous ones, though, next year's workshop will not
gather only invited researchers, but will be open to participants
submitting a contribution. Wishing to preserve the friendly and
cheerful atmosphere that characterized the series, we will limit the
number of participants to 50, and will achieve a balance between
invited talks and submitted contributions.
MOTIVATIONS
~~~~~~~~~~~
When natural language utterances are about sensible world, the
computation of the spatial and spatio-temporal reference plays a major
part in the construction of their formal representation. If the
understanding of a discourse is the ability to infer adequate answers
to questions about its informational content, the ability to deduce
properties of the discourse objects (like their localisation, their
structure or their shape) from the discourse representation, allows
the cognitive validation of these representations.
The most recent works in discourse theory (DRT, SDRT) clearly show the
necessity to take into account, in addition to linguistic and
pragmatic information, common knowledge about the universe of
discourse. In its whole generality, the formal representation of this
component of the meaning can very well be hopeless. We propose to
focus the attention on a specific category of discourses, namely
discourses which refer to the sensible world. In this case, common
knowledge reflects the structure and the properties of mental
representations of space, movement and time, these representations
being available not only through the analysis of linguistic
expressions but also through the analysis of different forms of
reasoning and decision-taking associated with perception.
TOPICS OF INTEREST
~~~~~~~~~~~~~~~~~~
We propose to discuss the possible contributions of spatial and
spatio-temporal knowledge representation and reasoning to discourse
interpretation; as well as the possible contributions of the analysis
of time, space and movement in language to the comprehension of the
organization of the perceived objects, and to the identification of
their cognitively relevant properties.
Contributions are invited on substantial and original research on
various aspects of time, space and movement, including, but not
limited to, the following.
A. Semantics of time, space and movement in natural language
- Lexical semantics : from linguistic and conceptual description
to formalisation
- From lexicon to sentence and discourse: role of the spatial and
spatio-temporal (S & ST) common-sense knowledge in discourse
interpretation
- Logics and deductive mechanisms:
* for the computation of the S & ST reference
* for the cognitive validation of discourse representations
B. Knowledge representation and S & ST reasoning
- Ontology of S & ST entities : philosophical analysis and
formalisation
- Mental representations of space, time and movement
- Mathematics of the sensible world
- Naive physics, qualitative S & ST reasoning
- Logics and visual reasoning
- Contributions to discourse representation
C. Relations between language and perception
- Imaginal and/or propositional structures of mental
representations
- From language to visual perception: from propositional to
numerical structures (image synthesis)
- From visual perception to language: from numerical to
propositional structures (image interpretation)
- Mathematical and logical problems of hybrid reasoning
INVITED SPEAKERS
~~~~~~~~~~~~~~~~
Nicholas Asher, Linguistics and Philosophy, Austin
Patrick Blackburn, Logic and Computational Linguistics,
Saarbruecken
Mimo Caenepeel*, Linguistics, Edinburgh
Anthony Cohn, Artificial Intelligence, Leeds
John Etchemendy, Philosophy, Stanford
Luis Farinas del Cerro, Logics and Computer Science, Toulouse
Christian Freksa*, Cognitive Science, Hamburg
Christopher Habel*, Cognitive Science, Hamburg
Patrick Hayes*, Artificial Intelligence, Urbana
Gerd Herzog, Artificial Intelligence, Saarbruecken
Hans Kamp*, Linguistics and Philosophy, Stuttgart
Manfred Krifka, Linguistics, Austin
Carlota Smith, Linguistics, Austin
Barbara Tversky*, Psychology, Stanford
Claude Vandeloise, Linguistics, Baton-Rouge
Achille Varzi, Philosophy, Trento
Henk Verkuyl, Linguistics, Utrecht
Co Vet, Linguistics, Groningen
(*) to be confirmed
PROGRAMME COMMITTEE
~~~~~~~~~~~~~~~~~~~
Chair : Mario Borillo, Artificial Intelligence, Toulouse
Nicholas Asher, Linguistics and Philosophy, Austin
Patrick Blackburn, Logics and Computational Linguistics, Saarbruecken
Andree Borillo, Linguistics, Toulouse
Anthony Cohn, Artificial Intelligence, Leeds
John Etchemendy, Philosophy, Stanford
Patrick Hayes, Artificial Intelligence, Urbana
Carlota Smith, Linguistics, Austin
Barbara Tversky, Psychology, Stanford
Achille Varzi, Philosophy, Trento
Co Vet, Linguistics, Groningen
Laure Vieu, Artificial Intelligence, Toulouse
FORMAT FOR SUBMISSION
~~~~~~~~~~~~~~~~~~~~~
Submitted papers should be at most 12 pages in length and be produced
in 12pt (default LaTeX article style is OK). Submissions should
provide the affiliation, full postal address, telephone and fax
numbers, and e-mail address (if any) of the author(s). A few words
stating the position of the paper with respect to the topics of
interest would be useful, as well as a 100-200 word abstract.
Electronic submission (plain ASCII, LaTeX, uuencoded PostScript, or
BinHex Mac Word files) is recommended. They should be sent to
tsm@irit.fr before 10 February 1995. Hard-copy submissions (4 copies)
should reach the Programme Chair no late than 10 February 1995.
Notification of acceptance will be sent to authors by 10 April, 1995,
and final versions (camera-ready) will be due by 15 May, 1995. These
will be compiled as Workshop Notes to be distributed to the
participants.
SCHEDULE
~~~~~~~~
Papers Submission............... 10 February, 1995
Notification of acceptance...... 10 April, 1995
Final version due............... 15 May, 1995
Workshop........................ 23-28 June, 1995
ORGANIZATION
~~~~~~~~~~~~
Organizing Committee:
Pascal Amsili, IRIT
Michel Aurnague, ERSS
Andree Borillo, ERSS
Mario Borillo, IRIT
Myriam Bras-Grivart, IRIT
Pierre Sablayrolles, IRIT
Laure Vieu, IRIT
Contact:
TSM'95
c/o Mario Borillo
IRIT - Universite Paul Sabatier
118, route de Narbonne,
F-31062 Toulouse Cedex
FRANCE
Tel: (+33) 61.55.60.91
Fax: (+33) 61.55.83.25
E-mail: tsm@irit.fr
WWW: http://www.irit.fr/ACTIVITES/EQ_LRC/tsm95.html
-----------------------------------------------------------------------
Date: Fri, 11 Nov 1994 06:49:37 +1000
From: sussex@lingua.cltr.uq.oz.au (Prof. Roly Sussex)
To: nl-kr@cs.rpi.edu
Subject: CFP: PACLING '95, Pacific Rim NLP Wkshps, Apr 95, Brisbane
CALL FOR PAPERS
(Third Circular and extended submission date)
* * * * * * * * *
PACLING '95
Pacific Association for Computational Linguistics
2nd Conference
April 19-22 (Wed-Sat) 1995
The University of Queensland
Brisbane, Queensland, Australia
* * * *
* HISTORY AND AIMS *
* * * *
PACLING (= Pacific Association for Computational
LINGuistics) has grown out of the very successful Japan-
Australia joint symposia on natural language processing
(NLP) held in November 1989 in Melbourne, Australia and in
October 1991 in Iizuka City, Japan. The first meeting of
the retitled PACLING, a name designed to express the wider
membership, took place in Vancouver, Canada in April 1993.
PACLING '95 will be a low-profile, high-quality, workshop-
oriented meeting whose aim is to promote friendly
scientific relations among Pacific Rim countries, with
emphasis on interdisciplinary scientific exchange showing
openness towards good research falling outside current
dominant "schools of thought," and on technological
transfer within the Pacific region. The conference is a
unique forum for scientific and technological exchange,
being smaller than ACL, COLING or Applied NLP, and also
more regional with extensive representation from the
Western Pacific (as well as the Eastern).
* *
* TOPICS *
* *
Original papers are invited on any topic in computational
linguistics (and strongly related areas) including (but not
limited to) the following:
Language subjects:
text, speech;
pragmatics, discourse, semantics, syntax,
lexicon, morphology, phonology, phonetics;
language and communication channels,
e.g., touch, movement, vision, sound;
language and input/output devices,
e.g., keyboards, menus, touch screens,
mice, light pens, graphics (incl. animation);
language and context,
e.g., from the subject domain, discourse,
spatial and temporal deixis.
Approaches and architectures:
computational linguistic,
multi-modal but natural-language centred;
formal, knowledge-based, statistical, connectionist;
dialogue, user, belief or other model-based;
parallel/serial processing
corpora and large-text linguistics
Applications:
text and message understanding and generation,
language translation and translation aids,
language learning and learning aids;
question-answering systems and interfaces to multi-
media databases
(text, audio/video, (geo)graphic);
terminals for Asian and other languages,
user interfaces;
natural language-based software.
* * * * ****
* SUBMISSION OF PAPERS *
* * * * ****
Authors should prepare full papers, in English, not more than
5000 words including references, approximately 20 double-
spaced pages. The title page must include: author's name,
postal address, e-mail address (if applicable), telephone
and fax numbers; a brief 100-200 word summary; and some key
words for classifying the submission.
Please send four (4) copies of each submission to:
Christian Matthiessen
Department of Linguistics
University of Sydney
Sydney 2006
AUSTRALIA
tel: +61 2 692 4227
fax: +61 2 552 1683
email: cmathies@extro.ucc.su.oz.au
* * **
* SCHEDULE *
* * **
NB NB NB NB NB NB NB NB NB NB NB NB NB NB NB NB NB NB NB NB
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
EXTENDED DEADLINE
+++++++++++++++++
In order to overcome some problems of information distribution,
the deadline for the submission of papers has been extended by
4 weeks.
Submission deadline: November 30th, 1994
Notification of acceptance: January 16th, 1995
Camera-ready copy due: March 1st, 1995
* * * * * * * * ****
* BACK TO BACK COGNITIVE SCIENCE CONFERENCE*
* * * * * * * * ****
The annual meeting of the Australian Cognitive Science Society
will be held at the University of Queensland from April 18-20.
Arrangements are being made with the organizers of that conference
for Pacling delegates to attend sessions of the Cognitive Science
Conference, and we are negotiating with them about joint sessions
and a coordinated program.
On the afternoon of Thursday 20 April CogSci95 will hold
a workshop on the applications of cognitive science. Pacling
delegates who would like to demonstrate software, or who would
like to offer a contribution, should contact Roly Sussex as soon
as possible.
* * * * * * *
* CONFERENCE COMMITTEE CHAIR *
* * * * * *
The Conference Committee Chair of PACLING'95 is
Roland Sussex
Centre for Language Teaching and Research
The University of Queensland
Queensland 4072
Australia
telephone: +61 7 365 6896
fax: +61 7 365 7077
email: sussex@lingua.cltr.uq.oz.au
* * * * * * * *
* PUBLICITY AND LOCAL ARRANGEMENTS *
* * * * * * * *
The conference will take place at the Centre for Language
Teaching and Research of the University of Queensland in
Brisbane, Australia. We are negotiating preferential rates
from downtown hotels.
Delegates may wish to visit attractions like the Barrier
Reef, Australia's desert centre or tropical rain forests
before or after the Conference, and we shall be negotiating
with travel companies to provide tour and travel
information.
For further information on the conference and on local
arrangements, contact
Hongliang Qiao
Centre for Language Teaching and Research
The University of Queensland
Queensland 4072
Australia
tel: +61 7 365 6897
fax: +61 7 365 7077
email: qiao@lingua.cltr.uq.oz.au
* * * * * * * *
* PACLING '95 COMMITTEES *
* * * * * * * *
Organizing Committee
Chair:
Naoyuki Okada (Kyushu Institute of Technology, Japan)
Members:
Naoyuki Okada (Kyushu Institute of Technology, Japan)
Christian Matthiessen (University of Sydney, Australia)*
Nick Cercone (Simon Fraser University, Canada)
Charles Fillmore(University of California, Berkeley, USA)
Conference Committee:
Chair:
Roland Sussex (University of Queensland, Australia)
Members:
Dan Fass(Simon Fraser University, Canada)
Randy Goebel(University of Alberta, Canada)
Kiyoshi Kogure(NTT, Japan)*
Paul McFetridge(Simon Fraser University, Canada)
Jun-ichi Nakamura(Kyushu Institute of Technology, Japan)
Minako O'Hagan(Victoria University, Wellington, New Zealand)
Fred Popowich(Simon Fraser University, Canada)
Hiroshi Sakaki(KDD, Japan)
Akira Shimazu (NTT, Japan)
Stanley Starosta(University of Hawaii, USA)*
Roland Sussex(University of Queensland, Australia)
Masami Suzuki(KDD, Japan)
* Program Coordinator
-----------------------------------------------------------------------
To: comp-ai-nlang-know-rep@uunet.uu.net
From: lucja@iwanska.cs.wayne.edu (Lucja Iwanska)
Subject: Announcement: Wkshp on context (NLP, KR, etc.), Aug 95, Montreal
Date: 11 Nov 1994 00:43:17 GMT
We would like to invite you to participate in the workshop
"Context in Natural Language Processing"
(see the attached tentative cfp with technical issues and agenda)
which, if approved by the IJCAI-95, will be held August 19-21, 1995,
immediately prior to the start of the main conference, Montreal,
Canada. All workshop participants are expected to register for the
main conference. Preliminary versions of papers will be due by 3.15.95.
The number of available slots is limited, and if you consider to
participate, please, send (preferrably by email) a tentative
confirmation in the following form
1. Name
2. Email address
3. Affiliation
4. One of the two:
(a) paper-and-attendance
(b) attendance-only
to Lucja Iwanska
Department of Computer Science
Wayne State University
Detroit, MI 48202
email: lucja@cs.wayne.edu
ph. (313) 577-1667
secr. (313) 577-2478
fx. (313) 577-6868
- - - - - - - - - - - - - -
IJCAI-95 Workshop (pending approval)
August 19-21, 1995
Montreal, Canada
CONTEXT IN NATURAL LANGUAGE PROCESSING
Correct interpretation of natural language utterances and texts
requires linguistic and non-linguistic context. The goal of this
workshop is to investigate the nature of context in natural language,
its role in natural language processing, and shed some light on this
largely unexplored area of great theoretical and practical importance.
Dialogue and text processing are two application domains where the
lack of good theories of context impedes significant progress in
applying and developing new technologies.
As speech technology matures, it becomes technically feasible to build
dialogue systems. However, understanding dialogues, and especially
multimodal dialogues, is not possible without some account of the role
of context. Similarly, with today's text processing technology it is
feasible to automatically create knowledge bases from fairly
unconstrained texts such as newspaper archives. Ignoring
context in such texts, however, results in knowledge bases that are
not only very incomplete, but also dramatically different from
knowledge bases created by humans, based on the same texts.
We invite papers from researchers active in the fields of natural
language processing, knowledge representation, and other related areas
addressing theoretical aspects of context and their implications for
designing practical NLP systems. We are interested in reports on
implemented NLP systems utilizing contextual information. We are also
interested in knowledge representation systems, inference methods, and
algorithms that would allow one to computationally handle specific
aspects of context.
PRELIMINARY AGENDA:
Our workshop will provide answers and insights into how to go about
answering a number of questions, including the following:
I: ROLE OF CONTEXT IN NATURAL LANGUAGE
What is context ?
What is
``context of the previous utterance/sentence''
``context of the dialog-so-far/text-so-far''
and what is the relationship between them ?
How many different contexts are there ?
What makes two contexts different ?
What is the relationship between formalization of context and
natural language ideas of context ?
What is the status of context in a formal representation
aiming at truthfully capturing all the characteristics
of natural language ?
Is context an inherent characteristic of natural language
that ultimately decides the formal power of natural language ?
Is natural language minus context a less powerful formal language ?
Does representing context and truthfully capturing characteristics
of natural language require new knowledge representation or automated
reasoning systems ?
What is the relationship between context,
and the semantics and pragmatics of natural language ?
Is context different from possible worlds and situations ?
What is the relationship between domain ontologies and contexts ?
II: CONTEXT-DEPENDENT INTERPRETATION OF NATURAL LANGUAGE
In which way does context affect interpretation of
natural language utterances and texts ?
Which aspects of context or which contexts result in
refined, more general, and different interpretations of natural language ?
Which aspects of context are explicit and which are implicit
in natural language utterances and texts ?
Which phenomena and inferences observed in natural language are
context-independent and which ones always depend on context ?
III: COMPUTABILITY
Does handling context increase or decrease computational complexity
of natural language processing ?
How to automatically identify context-provided constraints
resulting in conveying and understanding additional or different
aspects of information ?
How to represent those extra constraints provided by context
and how to automatically compute context-dependent
interpretation of natural language ?
Should the final interpretation of natural language be decontextualized
when stored in a knowledge base ?
How can information obtained in one context be utilized
in another, possibly unanticipated context ?
- - - - - - - - - - - - - -
TENTATIVE FORMAT:
In order to facilitate interaction and focus the discussion, we will
provide all the participants with specific examples and data
illustrating various aspects of context two months before the
workshop.
We will hold four sessions:
I: Role of context in natural language
II: Context-dependent interpretation of natural language
III: Computability
IV: General discussion
Sessions I, II, and III will be mildly structured:
First,
the committee will present a brief overview of possible answers
to the specific questions included in the agenda, and discuss their
own answers. This presentation will be mixed with questions
from the participants.
Second,
the participants whose papers were accepted will briefly comment
on their own answers to these questions. These presentations will
also be mixed with questions from the participants.
Session IV, a general discussion, will give each participant
a chance to make a statement about any context-related issue,
make a comment, raise a question, argue for or against some answers etc.
ORGANIZING COMMITTEE:
Keith Devlin
Saint Mary's College
R.V. Guha
Microelectronics and Computer Technology Corporation
Lucja Iwanska (Primary contact)
Wayne State University
Karen Jensen
Microsoft Corporation
John McCarthy
Stanford University
John F. Sowa
SUNY at Binghamton
Wlodek Zadrozny
IBM TJ Watson Research Center
-----------------------------------------------------------------------
From: hadley@cs.sfu.ca (Bob Hadley)
Subject: Announcement: Report on Connectionism, Semantics
Date: Mon, 14 Nov 1994 18:26:59 GMT
(Extended Abstract)
Strong Semantic Systematicity
from
Unsupervised Connectionist Learning
by
Robert F. Hadley and Michael Hayward
School of Computing Science
Simon Fraser University
CSS-IS TR94-02
Fodor's and Pylyshyn's arguments (1988) to the effect that human
thought and language exhibit both compositionality and
systematicity are by now widely known. Although connectionists
have questioned whether humans display these attributes in the
form that F&P describe (cf. van Gelder & Niklasson, 1994), most
now agree that in some important sense humans do employ a
combinatorial syntax and semantics and, as a result, exhibit some
form of linguistic systematicity.
In 1989--90, a number of connectionists reported results which
established that connectionist networks (hereafter, c-nets) could
exhibit forms of linguistic generalization, which,
prima facie, qualify as systematicity. These results were
obtained without recourse to mere implementation of ``classical''
symbolic methods, and so, it appeared that one of F&P's major
conclusions was falsified. However, in Hadley, 1992, 1994a, a
learning based conception of systematicity was introduced, and
various degrees of systematicity were distinguished, ranging from
weak syntactic to strong semantic systematicity. Hadley (1994a)
examined six different connectionist systems (Chalmers, 1990;
Elman, 1990; McClelland & Kawamoto, 1986; Pollack, 1990;
Smolensky, 1990; St.John & McClelland, 1990) and argued that,
in all probability, none of these systems displayed the strong
forms of systematicity that humans display. As a consequence,
it appeared that a variant of F&P's original challenge stood
unscathed. Recently, however, some researchers claim to have
satisfied Hadley's definition of strong systematicity, though not
his formulation of semantic systematicity. In one instance
(Phillips, 1994), this claim clearly requires qualification,
since (as Phillips has acknowledged, personal communication) the
system involved cannot process embedded sentences as required by
Hadley's definition. In another instance (Christiansen &
Chater, 1994), a claim to strong generalization is restricted to
a single syntactic context (conjunctive noun phrases).
Discussion of this claim, together with those of Niklasson & van
Gelder (1994) are given in Hadley, 1994b, where reservations are
explored. In any event, none of the researchers just cited
address semantic aspects of systematicity and compositionality,
although F&P's (1988) presentation of these concepts did seem to
involve semantic issues (such as the capacity to understand the
{\em meaning} of novel sentences and the need to banish semantic
equivocation in logical inference).
................................................
A network exhibits semantic systematicity just in case, as
a result of training, it can assign appropriate meaning
representations to simple and embedded sentences which contain
words in syntactic positions they did not occupy during training.
Herein we describe a network which displays strong
semantic systematicity in response to *unsupervised*
training. In addition, the network generalizes to novel levels
of embedding. Successful training requires a corpus of about
1000 sentences, and network training is quite rapid. The
architecture and learning algorithms are purely connectionist,
but `classical' insights are discernible in one respect, viz.,
that complex semantic representations spatially contain their
semantic constituents. However, in other important respects,
representations are distinctly non-classical.
The above is available as a PS file by email.
Also available by FTP upon request. Please contact
hadley@cs.sfu.ca for FTP directions.
End of NL-KR Digest
*******************
