[Prev][Next][Index]
Newsletter Jan. 31, No. 14
-
Subject: Newsletter Jan. 31, No. 14
-
From: csli@csli.stanford.edu
-
Date: Wed 30 Jan 1985 17:26:40-PST
C S L I N E W S L E T T E R
___________________________________________________________________________
January 31, 1985 Stanford Vol. 2, No. 14
___________________________________________________________________________
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, January 31, 1985
12 noon TINLunch
Ventura Hall ``Psycholinguistic Correlates of Defense and Coping''
Conference Room Hans Steiner, M.D.
Dept. of Psychiatry and Behavioral Sciences
2:15 p.m. CSLI Seminar
Redwood Hall ``An Internal Semantics for Modal Logic''
Room G-19 Moshe Vardi, CSLI
Discussion led by John Etchemendy
3:30 p.m. Tea
Ventura Hall
4:15 p.m. CSLI Colloquium
Redwood Hall No colloquium scheduled for this week
_________________
CSLI ACTIVITIES FOR *NEXT* THURSDAY, February 7, 1985
12 noon TINLunch
Ventura Hall Excerpts from Charles Bigelow's ``Principles of
Conference Room Structured Font Design for the Personal Workstation''
and Fernand Baudin's
``Typography: Evolution + Revolution''
Discussion led by David Levy
(Abstract on page 2)
2:15 p.m. CSLI Seminar
Redwood Hall ``Reasoning About Actions and Processes''
Room G-19 Michael Georgeff, CSLI
(Abstract on page 2)
3:30 p.m. Tea
Ventura Hall
4:15 p.m. CSLI Colloquium (See the notice below)
``From Sound to Score: Computerized
Transcription of Music''
Bernard Mont-Renard,
Center for Computer Research in Music and Acoustics
_________________
COLLOQUIUM NOTICE
Since next week's colloquium will not be held at Redwood Hall, it can only
accommodate 25 people and, unfortunately, is already full. If you wish to be
placed on the alternates list, please contact Ingrid Deiwiks (Ingrid@Turing,
497-3084).
Page 2 CSLI Newsletter January 31, 1985
___________________________________________________________________________
NEW CSLI REPORT
A new CSLI Report by Jon Barwise, ``The Situation in Logic--II:
Conditionals and Conditional Information'' (Report No. CSLI--85--21), has
been published. To obtain a copy of this report write to Dikran
Karagueuzian, CSLI, Ventura Hall, Stanford 94305 or send net mail to Dikran
at SU-CSLI.
_________________
ABSTRACT OF NEXT WEEK'S TINLUNCH
The TINlunch of February 7 will focus on some of the issues surrounding the
new computer technology exemplified by TEDIT, TEX, and EMACS. These ``word
processing'' and ``document preparation'' systems are, of course, nothing
other than ``writing'' tools -- intended for writing with the aid of the
computer. The first reading, an excerpt from an article by Charles
Bigelow, discusses the design of typefaces in the new digital medium as a
problem of balancing conservation and innovation: conserving the legibility
and elegance of our inherited letter forms while meeting the demands of the
new medium. In the second reading, Fernand Baudin suggests that the new
writing technology will require of us a new literacy: not just the ability
to read and write, but the ability to organize our writing visually -- that
is, typographically. He calls for ``the close cooperation of specialists
in many branches: linguist[ic]s, communication, psychology, history,
technology.'' --David Levy
_________________
ABSTRACT OF NEXT WEEK'S SEMINAR
``Reasoning About Actions and Processes''
Active intelligent systems need to be able to represent and reason about
actions and how those actions can be combined to achieve given goals. For
example, knowledge about kicking a football, performing a certain dance
movement, cooking a roast dinner, solving Rubik's cube, or diagnosing an
engine malfunction, is primarily knowledge about sequences of actions or
procedures for achieving these goals. Within AI, there have been two
approaches to this problem, with a somewhat poor connection between the
two. In the first category, there is some work on theories of action, or
what an action is. This research has focused mainly on problems in natural
language understanding concerned with the meaning of action sentences.
Second, there is work on planning, i.e., the problem of constructing a plan
by searching for a sequence of actions that yields a given goal.
Surprisingly, there is almost no work in AI about the execution of
pre-formed plans -- yet this is the almost universal way in which humans go
about their day-to-day tasks, and probably the only way other animals do
so. In this talk we aim to set the foundation for a theory of action that:
(1) provides a suitable semantics for simple action sentences in natural
language, (2) provides a method of practical reasoning about how to achieve
given goals based on procedural knowledge, and (3) serves as a basis for
planning. The first of these aims is met by defining a suitable
declarative semantics for action, and the second by providing a suitable
operational semantics. The third rests on both of these, but in addition
requires that we have a means of searching the space of possible world
histories. --Michael Georgeff
Page 3 CSLI Newsletter January 31, 1985
___________________________________________________________________________
SUMMARY OF THE F4 MEETING ON JANUARY 7
The topic was an overview belief revision as a research area in AI.
``Belief revision'' is a broad enough term to cover many different types of
inferential activity in AI. We discussed four types: (1) Search theory, in
which assumptions are made and retracted in an effort to find a problem
solution; (2) ``Truth'' maintenance systems a la Doyle. There are
foundational theories of belief in the sense of Harmon, with a set of
unsupported premises underlying all beliefs. The key feature of these
systems is their attempt to keep track of all justifications for belief,
and to revise these justifications in the face of contradictory belief.
(3) Database updates in the presence of integrity constraints or
user-defined views, in which case the update can become ambiguous. The
syntactic approach of Vardi et. al. was reviewed. (4) Ad-hoc approaches
designed for particular domains, for example the simple ``believe what you
see'' principle embedded in Shakey the robot.
Ned Block made the interesting observation that belief revision in the
AI context did not correspond to scientific theory revision as discussed in
the philosophical literature; for example, the principle of simplicity did
not seem to be a criterion for revision. This provoked a large amount of
discussion. --Kurt Konolige
_________________
SUMMARY OF THIS WEEK'S LOGIC SEMINAR
``Types and Partial Functions''
Gordon Plotkin
We presented a variant of Scott's semantic theory which makes use of
partial functions, avoiding the use of the bottom element to denote lack of
termination. There is a metalanguage for semantics with a compatible
operational and denotational semantics (at all types) where termination
corresponds to definedness. There is a logic, a variant of LCF, Scott's
Logic of Computable Functions; because of partiality this logic is free.
Claimed advantages of the system include its simplicity and naturality
(avoidance of some coding and correspondence with computational intuition).
_________________
ABSTRACT OF NEXT WEEK'S LOGIC SEMINAR
``What Does it Mean for Rewrite Rules to be ``Correct''?''
Feb. 4, 1985, 4:15, Faculty Lounge, Math Corner
Dr. Ed Wimmers, IBM, San Jose
We consider an operational definition for FP via rewrite rules. What would
it mean for such a definition to be correct? We develop a new formal
criterion for deciding whether there are enough rewrite rules and show that
our rewrite rules meet that criterion. Our proof technique is novel in the
way we use the semantic domain to guide an assignment of types to the
untyped language FP; this allows us to adopt powerful techniques from the
typed lambda-calculus theory.
_________________
NL4 MEETING
Tuesday, February 5, at 3:00 PM, Ventura seminar room
Last week, Phil Cohen presented the beginnings of a formalism for rational
interaction, upon which will be erected (in future meetings) a theory of
illocutionary acts. A simplified theory of rational action was proposed in
which it was deducible that a competent, rational agent, equipped with a
persistent goal and a correct plan (for the initial circumstances) would
either eventually achieve that goal, or would believe there is no longer
anything he could do to achieve it. Next week , definitions of requesting,
informing, and questioning will be derived, merging the above
characterization of action with a Gricean theory of communication.
Page 4 CSLI Newsletter January 31, 1985
___________________________________________________________________________
ABSTRACT OF NEXT WEEK'S NL1/NL3 MEETING
``The Definiteness Effect and Superlatives''
Anna Szabolcsi, MIT Center for Cognitive Science
February 6, Noon, Ventura Trailer Classroom
First, it will be argued that the distinction in Hungarian between prefixed
and non-prefixed verbs facilitates the identification of the definiteness
effect in a much wider range of contexts than those of ``there-insertion''.
The fact that all the relevant verbs act as bleached existential predicates
supports the semantic explanations of the definiteness effect. Second, in
the environment of local WH/FOCUS movement superlatives will be shown to
have a ``comparative'' reading on which they act as canonical indefinites.
The syntactic restrictions on the distribution and interpretation of
``comparative superlatives'' will be pointed out.
_________________
LINGUISTICS COLLOQUIUM
``How do Indexicals Fit into Situations? On Deixis in English and Polish''
Barbara Kryk, Adam Mickiewicz University, Poznan
Bldg. 200 (History Corner), Rm. 217
Tuesday, February 6, 3:15 pm
The task of describing deixis is not an easy one, since the notion in
question concerns a domain where language and reality meet. Pointing by
means of indexical expressions is, on the one hand, one of the chief
functions played by language in the communication process. On the other
hand, deixis is a source of such complex linguistic phenomena as reference
and anaphora. Hence, in analyzing deixis, one faces a dilemma: whether to
do it formally, within the framework of a truth-conditional semantics, or,
alternatively, to employ a pragmatic framework. The present study is an
attempt to reconcile the two possible ways of approaching indexicals. The
analysis is limited, for brevity's sake, to the demonstrative pronouns,
contrasting `this' and `that' with their Polish equivalents. The
distribution of proximal/distal demonstrative pronouns in Polish is far
from regular. The opposition often gets neutralized; only one pronominal
form is used to cover all cases (namely, the proximal variant `ten').
Since these facts are difficult to capture by means of pragmatic formulas,
a more formal framework is required, and it is Situation Semantics that
seems to be able to accommodate the adduced data most satisfactorily.
Since demonstrative pronouns have been traditionally considered among the
most basic indexicals, it is suggested that the observations made
concerning them might be of some more universal value, as they could be
extended onto demonstrative adverbs and other deictic expressions as well.
_________________
TUTORIAL LECTURES ON LINGUISTICS FOR NON-LINGUISTS
The second of three sessions intended for non-linguists on Government and
Binding (Chomsky), Generalized Phrase Structure Grammar (Gazdar/Klein/
Pullum/Sag), and Lexical Functional Grammar (Bresnan/Kaplan) takes place on
February 5. The sessions are in the trailer seminar room on Tuesdays from
1:00 to 3:00 PM. Each lecturer sketches the leading ideas of one theory,
and (perhaps) contrast it with the other theories. The complete schedule
is listed below.
January 29 GB Lecture by Peter Sells
February 5 GPSG Lecture by Geoff Pullum
February 12 LFG Lecture by Joan Bresnan
Linguists are asked not to attend.
-------