CCAT: TIME: Fantasyland?/Various issues
fritz@rodin.wustl.edu (Fritz Lehmann)
Date: Mon, 10 Oct 94 20:55:48 CDT
From: fritz@rodin.wustl.edu (Fritz Lehmann)
Message-id: <9410110155.AA17931@rodin.wustl.edu>
To: cg@cs.umn.edu, phayes@cs.uiuc.edu, srkb@cs.umbc.edu
Subject: CCAT: TIME: Fantasyland?/Various issues
Cc: M.J.Johnson@qmw.ac.uk, anquetil@IRO.UMontreal.CA,
beancar@cucis.cis.columbia.edu, bill@violin.att.com,
billrich@vnet.ibm.com, brayman@zuben.ca.boeing.com,
buteau_brandon@prc.com, cassidy@micra.com, cbwillis@netcom.com,
cyre@vtvm1.cc.vt.edu, dick@Glue.umd.edu, doudna@aol.com,
doug@csi.uottawa.ca, fletcher.mcleancsd@xerox.com,
fritz@rodin.wustl.edu, ged@cs.rmit.edu.au, gerbe-o@immedia.ca,
grau@falcon.depaul.edu, kivs@bgcict.bitnet, kremer@cpsc.ucalgary.ca,
kschoi@cs.kaist.ac.kr, lukose@peirce.une.edu.au, moulin@ift.ulaval.ca,
oh@vax2.cstp.umkc.edu, peterman@informatik.uni-hamburg.de,
roger@ci.deere.com, s.griffin@mcs.surrey.ac.uk,
shmyaeng@mailbox.syr.edu, sowa@turing.pacss.binghamton.edu,
thompson@zuben.boeing.com, wei@oucsace.cs.ohiou.edu, willems@cs.vu.nl
Sender: owner-srkb@cs.umbc.edu
Precedence: bulk
On the CCAT/TIME deep time ontology, Pat Hayes wrote:
>> Pat, can your three time ontologies handle all of the above?
>>Can they bear the strain?
>Of course not. This is several major research programs you have so casually
>outlined here. I havent yet seen a satisfactory account of imprecision and
>tolerance, for a start, including a few of my own. And "everything required
>by natural language"?? Isnt that just everything?
>Be a little more realistic, Fritz, or we are all in fantasyland from the
>very beginning.
I hope not. The mission of the CCAT TIME subgroup is to proffer an
ontology of time which others can use for general purposes. If that is
"fantasyland" then the effort is doomed from the start. In fact your
system sounds pretty close to being a good basis for what is needed, and
it does things that never occurred to me. Such an ontology can be
"rich" and somewhat deep, but we need not ground everything in the
ultimate depths. The base ontology doesn't need to _do_ everything, but
it should _allow_ everything (of likely practical value). Thus pure
TIME should not deal with tenses, or indexicality, but other people such
as Bernard Moulin who are creating tense systems should be able to FTP
the TIME ontology and use it for their ends. Even the dates and times I
mentioned before are really a combination of TIME with MEASUREMENT UNITS
rather than being basic to TIME.
We will definitely need one or more conventions for imprecision and
tolerance (for many areas, not just TIME). The DARPA Rome Planning
Ontology has a scheme for imprecise time. Also there are books on this
kind of thing in measurement theory. It's going to be highly context-
(purpose)-sensitive in common-sense reasoning. I'm not sure that the
essentials of this should have any special connection to TIME. The
annoying thing is that, on the one hand, turning temporal order
relations into tolerance relations loses logical transitivity, but on
the other hand we in fact assume and use transitivity for the common-
sense versions of these relations except in extreme borderline
("pathological") cases.
>I wanted actual concepts. For example, is the idea of an
>intermittent interval ("every wednesday afteroon for the next five weeks")
>a very important one for people?
It is something that a general-purpose time system should be able
to say. Again, that doesn't mean we have to define this case, but the
language should allow defining it. I suspect that "intermittent
intervals" can just be put together from unbroken intervals using
negation for OFF periods and conjoining interval assertions. Right? No
need to provide arbitrary algebraic interval structures, prime
intervals, Cantor Dust, etc.
If I correctly understood an earlier remark of yours, then the
Rational time-line would be the preferred default model rather than the
Reals or a fine discrete chain of Integers.
>For the record, my theory (actually a collection of alternative theories)
>can (like many others) handle standard time durations of various kinds and
>describe clocks and calendars with various corrections, including regular
>corrections like leap days. It can handle notions of faster or slower, and
>what I call 'compound clocks', ie clocks which simultaneously measure time
>on several scales. All binary relations between intervals in linear time,
>including Allen's six, are definable. It supports several different notions
>of timepoint and timeinterval and the relationships between them, and also
>can describe things I call 'moments' which are pointlike intervals. Time
>can be dense or discrete or any sensible combination, and it can branch or
>not (usually not).
Sounds pretty good. When you are in the home stretch (or else
now), please send the CCAT list a bare list of your available TIME
predicates and relations (both primitive and defined). Your explicit
inclusion of "clocks" in TIME is an interesting decision; I'll be
curious to see which time predicates depend on the existence of clocks.
See my email today answering Allen Brewer's question about a deadline's
effect on STATE. I suspect that I will push for including some tedious
subtleties, such as the role of authority in determining time standards.
>So far I havnt described geographical variation as in
>things like 'GMT', although that would be easy given some other
>axiomatisation of the notion of location. Sidereal time might be a little
>trickier, but I bet it could be arranged.
These should be done in conjunction with the CCAT SPACE/GEOGRAPHY
subgroup using the predicates and relations of those subjects. Would
someone propose what should be available for this purpose from the
SPACE/GEOGRAPHY ontology? In fact, who is actually going to do the
SPACE/GEOGRAPHY ontology?
>I havnt thought about relativistic time, and the theory, like any logical
>axiomatisation, cannot handle indexicals. (The only indexical in KIF means
>'this database', which isnt sufficient to allow temporal change.)
I thought KIF dropped that indexical "THIS-DATABASE" idea.
Have they restored it?
>Question:
>Is it really important to be able to represent the conceptual framework
>behind these ideas, or is it sufficient to represent simply the ideas? Must
>a temporal reasoner know relativity theory, for example?
>Pat
Since the Global Positioning System (GPS) requires relativistic
adjustments, and it has considerable practical value, I think it's worth
flagging the intial CCAT time ontology as "NON-RELATIVISTIC" and leave
it at that. That way somebody (not I) someday may be able extend the
time ontology to include relativity without causing incompatability
troubles with pre-existing knowledge bases.
----------------------------------------------------------------------------
Pat Hayes further asked, in a separate message:
----begin quote----
Why do you bracket OBJECT with EVENT and PROCESS? (Are these all the things
that can have or be in a STATE?) And what is your distinction between an
EVENT and a PROCESS? (This one has been made many times, often with subtly
different meanings.)
---end quote----
These are in the same Peirce CCAT ontology subgroup OBJECT/EVENT/PROCESS
because we will need a good definition of the differences among them.
Is a thunderstorm an event, an object or a process? Etc. etc. We
need a bit of deep thought on this. Lenat's CYC calls people processes.
The Penman Upper Model requires the distinction.
Before we can nail down a base-set of primitive case-like
relations (between an event and its participants) we may have to decide
what the underlying ontology of events is. As I intimated before, the
_words_ "event", "process", "object", "continuant", "occurrent", "fluent",
"act", "action", etc. could cause needless squabbles -- I'd prefer to
stick to our own officially defined versions with sets of real-world
exemplars and non-exemplars. (EVENTOIDS?) A system
of event-state-event sequences works nicely in some domains, like
idealized electronic devices or particle systems, but not so
neatly in areas like internal medicine, business or weather. You say the
falling of a ball, while it is falling, is not an event. Is its passing a
boundary an event? Arrival of the millenium will not be an event to me, but
some others would disagree. I'd like primitives for the common-sense
notion of an event and its participating objects. I think we may have to
recognize explicitly that we use a shorthand (OBJECT/EVENT/PROCESS)
to describe at the macro level what at the underlying physics level
is just a seething mass of trajectories. I don't know offhand how to
put that fact into the ontology.
Beside me is a bunch of unread books pertaining to this issue,
including among others Process and Reality by Whitehead, Events in the
Semantics of English by Parsons, Action Semantics by Mosses,
The Temporal Logic of Reactive and Concurrent Systems by Manna
and Pnueli, Model Based Systems Engineering by Wymore, Intending
and Acting by Brand, Naive Semantics for Natural Language
Understanding by Dahlgren, Multifactted Modelling and Discrete
Event Simulation by Ziegler, Causal Necessity by Skyrms, Roget, etc.
Hasn't somebody out there in email-land already studied this stuff?
I hope someone can opine on all this for the benefit of CCAT,
(Maybe I won't have to read all those books any time soon).
Also, does anyone know what the current event-system of Schubert &
Hwang's Episodic Logic is?
Hayes says the distinction between EVENT and PROCESS "has
been made many times, often with subtly different meanings." Is
there an existing survey tersely comparing these? CCAT really
has to bite the bullet and decide on an ontology for this before
much else can be done.
Yours truly, Fritz Lehmann
GRANDAI Software, 4282 Sandburg Way, Irvine, CA 92715, U.S.A.
Tel:(714)-733-0566 Fax:(714)-733-0506 fritz@rodin.wustl.edu
=============================================================