Copy Link
Add to Bookmark
Report
NL-KR Digest Volume 02 No. 01
NL-KR Digest (1/05/87 11:22:13) Volume 2 Number 1
Today's Topics:
Seminar - Misunderstandings in Natural Conversation (BBN)
Conference - Program for TINLAP3
----------------------------------------------------------------------
Date: Sun, 4 Jan 87 14:29:29 EST
From: "Steven A. Swernofsky" <SASW%MX.LCS.MIT.EDU@MC.LCS.MIT.EDU>
Subject: Seminar - Misunderstandings in Natural Conversation (BBN)
TOPIC: MISUNDERSTANDINGS IN NATURAL CONVERSATION
SPEAKER: PROFESSOR EMANUEL A. SCHEGLOFF
DEPARTMENT OF SOCIOLOGY
UCLA
3 - 5 pm
Thursday, January 8, 1987
2nd floor large conference room
10 moulton street
Professor Schegloff, who is the most important researcher in the
Ethnomethodological School of "Conversational Analysis" will discuss work on
the source of misunderstandings in conversation. He will demonstrate how,
using C.A. techniques, a number of sources of misunderstandings can be
identified in everyday talk through a very careful analysis of the progress of
the talk subsequent to the site of difficulty.
The talk will be illustrated with examples taken from everyday conversational
interaction.
------------------------------
Date: Sat, 3 Jan 87 21:38:08 mst
From: yorick%nmsu.csnet@RELAY.CS.NET
Subject: Program for TINLAP3
Program for Tinlap3
Computing Research Laboratory
NMSU, Las Cruces, New Mexico,
January 7-9, 1987.
Tuesday January 6.
2-6pm. Registration. Lobby, Holiday Inn.
6-8pm. Reception. Double Eagle, Mesilla (tickets in registration packages).
Wednesday January 7.
(All sessions will be held in the Music Center Auditorium, NMSU)
8.30am. Panel I.
10.30am. Coffee
11am. Panel II.
1pm. Lunch
3pm Panel III.
6-8pm. Reception. Budaghers. (tickets only, limited accomodation)
Thursday January 8.
8.30am. Panel IV.
10.30am. Coffee
11am. Panel V.
1pm. Lunch
2.30pm. Panel VI.
Friday January 9.
8.30am. Panel VII.
10.30am. Coffee
11am. Panel VIII.
1pm. Lunch
2.30pm. Panel IX.
******
The Panels
Words and world representations.
How have these suddenly become more interesting? Do they offer a way through
from the old "primitive " dispute, and do they offer a way out from having to
separate world and linguistic knowledge? How does what we know about words fit
into the language understanding and generation process, and is that different
for understanding and generation ?
Don Walker (Bellcore) Chair
Bran Boguraev (Cambridge)
Bob Amsler (Bellcore)
Jerry Hobbs (SRI)
Judy Kegl (Princeton)
II. Unification and the new grammatism
How far does this really differ from the CFG position of the sixties? Does it
yet have any empirical successes in terms of working systems? To what extent
are these grammatical formalisms motivated by processing considerations ?
To what extent are these processing claims substantiated ? Are we converging
to some class of formalisms that are relevant for processing and, if so, how
can this class be characterized in a theoretical manner ?
What are the prospects of these types formalisms becoming the basis for future
natural language processing research ? Has the processing paradigm now really
fundamentally influenced linguistics ? Do processing considerations and
results show that such systems when implemented can be neutral between
analysis and production.
Has everyone really been doing unification for decades and just found out? Is
it a real advance or just a Hollywood term?
Fernando Pereira (SRI) Chair
Gerald Gazdar (Sussex)
Steve Pulman (Cambridge)
Aravind Joshi (U.Penn.)
Mitch Marcus (Bell Labs.)
Martin Kay (XEROX-PARC)
III. Connectionist and other parallel approaches to natural language processing
Is NLP inevitably commited to a symbolic form of representation? Can
syntactic, semantic or world knowledge be represented in that paradigm if
taken seriously? What parts of current CL will fare worst if there turn out to
be significant empirical advances with connectionist parsing? Are there any
yet (i.e. how far do we trust simulations programmed only on serial machines?)
What new approaches to syntax, semantics or pragmatics will be needed if this
approach turns out to be empirically justified? Will it just bring back all
the old views associated with associationism, and will they be changed in the
journey? Is parallel parsing just a new implementation or a real paradigm
shift?
Dave Waltz (Thinking Machines) Chair
Martin Kay (XEROX-PARC)
Gary Cottrell (UCSD)
Gene Charniak (Brown)
Jay McClelland (CMU)
Wendy Lehnert (U.Mass.)
IV. Discourse Theory and Speech Acts.
Is there yet any serious discourse theory with testable computational and
empirical consequences? What phenomena ought a processing theory of discourse
understanding/generation to address itself that are not already being attended
to currently? What aspects of discourse are language problems and which are
general AI/KR problems? What makes a theory of discourse a processing theory?
Does spoken language affect one's theory of discourse?
Is there any real hope that we will be able to recognise the plans/goals etc.
of a speaker? How much of conversation is carried on through the linguistic
window anyway? Do current theories of text and dialogue discourse mesh, and
should they?
Barbara Grosz (SRI) Chair
Julia Hirschberg (Bell Labs.)
Ray Perrault (SRI)
Bob Wilensky (Berkeley)
Franz Guenther (Tuebingen)
V. Why has theoretical NLP made so little progress?
Has CL advanced in this respect since Tinlap2 in 1978? What can NLP systems do
today in the light of what we would have predicted at Tinlap2. Why are we no
nearer to a common notation for systems since KRL----- would we be helped by
CL textbooks geared to particular programming languages (one such is now in
preparation)? Is it a case of just cycling through ranges of obscure
syntactic and semantic formalisms (and then rediscovering them every 10 years
or so)? Are there serious problems about the overall cognitive paradigm being
applied to NLP? Are there any serious alternatives to the current paradigms,
and what would they imply to NLP research directions and goals?
Roger Schank (Yale) Chair
Norm Sondheimer (USC-ISI)
Larry Birnbaum (Yale)
Ivan Sag (Stanford)
David Israel (SRI)
VI. Formal versus commonsense semantics.
What does Montague grammar or situation semantics have to say to CL? Can we
distinguish the good parts from what is bad and useless? For what NLP
applications might these formalisms be particularly appropriate? What have
such theories chosen to ignore, in terms of data or intuitions? How are they
to be computed: compositionally, randomly? How well can such formalisms mesh
with the rest of language representation processes, e.g. discourse and
pragmatic analysis?
Yorick Wilks (NMSU) Chair
David Israel (SRI)
Geoff Nunberg (Stanford/Xerox-PARC)
Wendy Lehnert (U.Mass)
Karen Sparck-Jones (Cambridge)
Susan Stucky (Stanford-CSLI)
VII. Reference: the interaction of language and the world.
When is a noun phrase a referring expression? How does the meaning of a noun
phrase contribute to the success of a referring act? How can a "wrong"
description be useful for referring? Is there any role for Russell's analysis
of descriptions in a pragmatic theory of referring?
What does it mean for a hearer to identify a referent? What is the
relationship between knowing who or what something is and referent
identification?
Is referring to events and situations inherently different from referring to
material objects? What identification criteria are applicable to events and
situations?
Doug Appelt (SRI) Chair
Deborah Dahl (SDC Inc.)
Bonnie Webber (U. Penn.)
Amichai Kronfeld (SRI)
Brad Goodman (BBN)
VIII. Metaphor
How relevant are the philosophical, linguistic, and psychological literatures
on metaphor? Can any of the recent work in dialogue, planning & speech acts
be applied to understanding metaphor. Are existing knowledge formalisms (e.g.
conceptual dependency, scripts, semantic networks, KLONE) adequate for
metaphor? If not, why not? Given that the recognition of metaphor involves
matching together large-scale knowledge structures, are there any existing
procedures that do this adequately? How can this matching be done? How might
we record the degree of match? Are there additional types of processing
necessary for recognising metaphor?
How should metaphor be represented in semantic representations of text? Are
there situations when a metaphor should be "resolved", and others when its
tension should remain? How can we recognise those situations?
Deirdre Gentner (Illinois) Chair
Andrew Ortony (Illinois)
Ed Plantinga (Toronto)
George Lakoff (UCB)
Geoff Nunberg (Stanford/Xerox-PARC)
IX. Natural Language Generation
Will the demands of language production bring AI, theoretical linguistics (and
of course CL) closer together than the demands of comprehension did in the
past? Is there anything special about generation?
Does generation constrain problems differently from understanding, in that it
would not matter if some high-powered machine could understand things no human
could say, but it would matter if the same machine generated them! Are
knowledge structures, of the world as much as language, the same or different
for understanding and generation? What is the relation between the message
the system wants to convey and its lexical, syntactic etc. abilities to do it.
Aravind Joshi (U. Penn) Chair
Dave MacDonald (U.Mass. Amherst)
Doug Appelt (SRI)
Bill Mann (USC-ISI)
Mitch Marcus (Bell Labs.)
Tony Kroch (U. Penn.)
------------------------------
End of NL-KR Digest
*******************
