Copy Link
Add to Bookmark
Report
NL-KR Digest Volume 02 No. 07
NL-KR Digest (2/10/87 17:44:01) Volume 2 Number 7
Today's Topics:
Seminar: Decidable, Logic-based KR (CMU)
Analysis of Syntactic Predictions (CMU)
From CSLI Calendar, Feb. 5, No. 15
"WHAT'S IN A WORD?" by Martin Kay
Seminar - Understanding Natural Language (SMU)
Generating NL Descriptions for Image Sequences
----------------------------------------------------------------------
Date: 6 Feb 87 12:13:33 EST
From: Marcella.Zaragoza@isl1.ri.cmu.edu
Subject: Seminar: Decidable, Logic-based KR (CMU)
TOPIC: Decidable, Logic-Based Knowledge Representation
SPEAKER: Peter Patel-Schneider, Schlumberger Palo Alto Research
PLACE: Wean Hall 5409
DATE: Tuesday, February 10, 1987
TIME: 3:30 pm
ABSTRACT
For a knowledge representation system to really represent knowledge
there must be some model-theoretic semantics for the information it
stores. The standard way of supplying such semantics is to use the semantics
of a formal logic, almost always one that is at least as strong as
standard first-order logic. Unfortunately, this results in a system that has
operations that are not guaranteed to terminate.
It is possible to give a complete characterization of expressively powerful,
but computationally tractable, knowledge representation systems if, instead
of being based on a standard logic, they are use a weaker logic with good,
model-theoretic semantics where reasoning is better behaved computationally.
Such systems are incomplete with respect to standard logics, but do have a
complete, semantic characterization of their reasoning.
The four-valued semantics of tautological entailment forms a starting point
for the investigation of such logics. A variant of first-order tautological
entailment, with tractable reasoning, is investigated. This logic is
extended by adding a terminological reasoner to create a stronger logic that
can be used as the basis of a knowledge representation system.
------------------------------
Date: 5 February 1987 1519-EST
From: Elaine.Atkinson@A.CS.CMU.EDU
Subject: Analysis of Syntactic Predictions (CMU)
Attention: cogsci bboard
Resent-Message-Id: <12277567511.14.LAWS@SRI-STRIPE.ARPA>
SPEAKER: Dr. Hiroyuki Musha, Visiting Scientist at the
International Center for Machine Translation
TITLE: "Analysis of Syntactic Predictions and its Application
to Parsing"
DATE: Tuesday, February 10
TIME: 12:00 - 1:20 p.m.
PLACE: Adamson Wing, Baker Hall
ABSTRACT: I will focus on syntactic predictions or expectations
of natural languages, and re-examine parsing methods of natural
languages from this viewpoint. At first I will classify
syntactic predictions into six types looking at examples for
each type of prediction. Then I will reconsider two parsing
methods proposed for natural languages from the viewpoint of the
classification of syntactic predictions. The first method is the
predictive analyzer proposed by Kuno in the 1960's, and the
second one is the LR parser which was also used recently for
parsing of natural languages.
As the name indicates, Kuno's predictive analyzer is based on
the analysis of predictions which is emitted while we read
English sentences. While it seems that people do not backtrack
so often when we read sentences, a lot of backtracking occurs
if we use this method to analyze sentences. I will explain
the reasons and necessary modifications based on the classification.
I will also present the results of an experiment in which about
300 sentences of four kinds of articles were input into the
modified parser.
The second parser, the LR parser, was originally developed for the
analysis of programming languages and is very efficient. This
parser, however, cannot handle some types of predictions naturally.
Modifications of the method for natural language parsing will be
discussed in the talk.
------------------------------
Date: Wed 4 Feb 87 17:01:37-PST
From: Emma Pease <Emma@CSLI.Stanford.EDU>
Subject: From CSLI Calendar, Feb. 5, No. 15
Tel: (415) 723-3561
[Excerpted from CSLI Calendar]
A Defense of Realism
John Perry
February 12
Throughout the history of philosophy there have been a number of
debates about "realism." The reality of a number of sorts of things
has been challenged and defended: universals, particulars, objects
outside the mind, objects inside the mind, minds, modal facts, any
kinds of fact, numbers, sets, other minds, one's own mind, fictional
objects, nonfictional objects, and so forth.
I shall say a few things about the structure of a typical
realism-debate, adopting the common-sense view that in some of these
debates the realists were right, and in others the antirealists were.
Then I shall move to consider a certain common antirealist
position. This position combines antirealism about the external
world with antirealism about properties and relations, although this
is often not so obvious because of the wide variety of things, other
than properties and relations, that people are willing to call
"properties" and "relations." The basic idea is the structure that
needs to be postulated to explain our mental and linguistic activities
is found in mind and language, not in an external world considered
apart from mind and language. This view is sometimes defended as
"methodology," sometimes as metaphysics.
Antirealism about the external world is popular among 19th century
philosophers and 20th century computer scientists. I shall argue that
the current favor it finds with (some) computer scientists is due to
their having not paid sufficient attention to various technical
distinctions in philosophy and logic. For example, computer
scientists often use the word "naive" as an empty epithet, so that
someone who is a realist is a "naive realist" to those who disagree
with him or her. Actually, "naive" as an adjective to "realist" is a
technical term, which contrasts with "critical," critical realism
being a position that at least attempts to take account of the
important contributions made by the structures of mind and language to
our interactions with the external world.
I shall pay particular attention to the works of Terry Winograd,
who has been led to antirealism conclusions from the same premises
that have led others (like Barwise and myself) to a form of critical
realism, and to a remark attributed to Moshe Vardi, namely, "If you
made a robot that could swim, it would swim in conceptual water,"
which apparently embodies the insights of a generation of computer
scientists, but remains pretty opaque to me.
------------------------------
Date: Fri, 6 Feb 87 19:59:12 PST
From: Subhana.pa@Xerox.COM
Subject: "WHAT'S IN A WORD?" by Martin Kay
Reply-To: Diebert.pa@Xerox.COM
Resent-Message-Id: <12277566663.14.LAWS@SRI-STRIPE.ARPA>
PARC FORUM
Thursday, February 12, 1987
4:00 pm, PARC Auditorium
MARTIN KAY
Principal Scientist, Natural Language Theory & Technology Area
Intelligent Systems Laboratory, Xerox PARC
will speak on
WHAT'S IN A WORD?
ABSTRACT: Recognizing words in English text is generally easy in
principle--break the text into contiguous sequences of alphabet
characters and look each of them up in an alphabetical list. That is in
principle. It practice, it works poorly in some fields, like chemistry
and medicine, in some languages, like German and Arabic, and for some
applications, like dictionaries and spelling checkers for small
machines. The problem is that the list could be large--potentially
infinite--in size, though the information it contains is tantalizingly
redundant. Obvious strategies are to subdue the problem with
computational cunning, mathematical magic, or linguistic lore. This
talk will apply all three of these to English "unredegaussability",
Spanish "digamelo", German
"Strassenbahnschafnerversicherungsgesellschaft", Arabic "atakaatab", and
Turkish "otobusun" among others, with the aim of showing that
computation, mathematics, and linguistics combine to make a truly
dynamic trio.
This Forum is OPEN. All are invited.
Host: Tim Diebert (Computer Science Laboratory, 494-4433)
Refreshments will be served at 3:45 pm; please don't come for
refreshments if you aren't attending the Forum!
Requests for videotaping should be sent to Susie Mulhern
<Mulhern:PA:Xerox or Mulhern.pa> before Tuesday noon.
Directions to PARC:
The PARC Auditorum is located at 3333 Coyote Hill Rd. in Palo Alto. We
are between Page Mill Road (west of Foothill Expressway) and Hillview
Avenue, in the Stanford Research Park. The easiest way here is to get
onto Page Mill Road and turn onto Coyote Hill Road. As you drive up
Coyote Hill, PARC is the only building on the left after you crest the
hill. Park in the large parking lot, and enter the auditorium at the
upper level of the building. (The auditorum entrance is located down
the stairs and to the left of the main doors.)
------------------------------
Date: 10 OCT 86 17:02:23 CDT
From: leff%smu@csnet-relay.arpa
Subject: Seminar - Understanding Natural Language (SMU)
An Overview of Understanding Natural Language
(The talk has already been given.)
The term natural language processing'' covers a broad range of topics
dealing with the use of human languages with computers. Areas of interest
include written language understanding and generation, spoken language
understanding and generation, machine translation between languages,
computer-aided study of linguistics, etc. This talk will concentrate on
understanding written language.
The first part of the talk will ask and partially answer the question
What constitutes a natural language and why is understanding one hard?''
Human languages are called natural to distinguish them from computer,
mathematical, and other languages. In reality, a formal human language
isn't really natural but is governed by a set of rules, albeit ambiguous
rules. We will look at why understanding written language is still very
hard.
The second part will look into some of the techniques used to attack the
problems raised in the first part. We will look at syntactic, semantic and
other approaches including ATNs, Conceptual Dependencies, modified LR
parsers, keyword matching, etc. The idea of asking the user for help
will also be presented.
Biography:
- B.S.C.S., University of South Florida, 1985.
- Master's student at SMU since June 1985. Principle work has been on
a symbolic math system for microcomputers and personal research on
natural language processing.
(Stephen Robinson, robinson%smu@csnet-relay)
------------------------------
Resent-Message-Id: <8702092205.AA11116@Cayuga.cs.Rochester.Edu>
Date: Mon 9 Feb 87 11:26:29-PST
From: Doug Appelt <APPELT@SRI-WARBUCKS.ARPA>
Subject: Generating NL Descriptions for Image Sequences
Prof. Wolfgang Wahlster of the Univeristy of Saarbruecken will
give a talk and demonstration of his systems on Friday February 20th
at 10 AM.
GENERATING NAUTRAL LANGUAGE DESCRIPTIONS FOR IMAGE SEQUENCES
Wolfgang Wahlster
Computer Science Department
Univerity of Saarbruecken
West Germany
The aim of the project VITRA (VIsual TRAnslator) is the
development of a computational theory of the relation between natural
language and vision. In this talk, we will focus on the semantics of
path prepositions (like "along" or "past") and their use for the
description of trajectories of moving objects, the intrinsic and deictic
use of spatial prepositions and the use of linguistic hedges to express
various degrees of applicability of spatial relations.
First, we describe the implementation of the system CITYTOUR,
a German question-answering system that simulates aspects of a
fictitious sightseeing tour through a city. Then we show how the
system was interfaced to an image sequence analysis system. From
the top of a 35m high building, a stationary TV camera recorded an
image sequence of a street crossing on video tape. In 130 selected
frames the moving objects were automatically recognized by analyzing
displacement vector fields. Our system then answers natural language
queries about the recognized events.
Finally, we discuss current extensions to the system for the
generation of a report on a soccer game that the system is watching.
Here we focus on the problem of incremental, real-time text generation
and the use of a re-representation component that models the assumed
imagination of the listener.
------------------------------
End of NL-KR Digest
*******************
