[Top] [All Lists]

Re: [ontolog-forum] Time representation

To: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Jerry Hobbs <hobbs@xxxxxxx>
Date: Mon, 21 Jan 2008 20:27:21 -0800
Message-id: <479570A9.5010108@xxxxxxx>
Sorry for the late response. Life outside ontolog-forum (shocking!)
and other similar excuses.    (01)

The current best pointer to OWL-Time is    (02)

    http://www.isi.edu/~hobbs/owl-time.html    (03)

In fact, the name "OWL-Time" is a bit of a historical accident.  It
was originally "DAML-Time", and done before it was really decided what
DAML was going to be.  After Ian Horrocks capitalized on the fact that
no one else was doing anything and got OWL accepted as the realization
of DAML, it was deemed that all "DAML-X" resources would be renamed to
"OWL-X".    (04)

The basic ontology is written in textbook logic.  We made an effort to
convert all that to CL, but I don't think we ever finished.  In any
case, it is trivial to do.  OWL is good for declaring predicates, and
for expressing a few kinds of simple axioms, but not much more.  It is
certainly not up to the very moderate complexity of an ontology of
time, as John Sowa pointed out.  Nevertheless, we wrote an OWL
ontology of whatever of the time ontology translated naturally into
OWL.  This is essentially a "projection" of the full ontology onto
OWL.    (05)

John Sowa:
 > In representing time, the first option is to choose between an
 > explicit time or an implicit "temporal logic", which does not
 > actually refer to time.
 > Temporal logics avoid references to time by using operators
 > such as 'sometimes', 'always', 'before', and 'after'.  That
 > representation, which was developed by Arthur Prior (a good
 > name for a philosopher who was writing about time), treats
 > time as a kind of modality (with 'always' corresponding
 > to 'necessity' and 'sometimes' to 'possibility').    (06)

OWL-Time is a theory of explicit times.  I agree with John's implicit
criticism; using temporal operators is inherently limiting in what can
be naturally expressed.  If you are going to introduce new operators
for concepts as simple as "always", you are going to end up with an
awful lot of operators before you are through.  Just a matter of
aesthetics, but I think the result would be extremely ugly.    (07)

John Sowa:
 > An approach with explicit time represents time with a linear
 > coordinate system.    (08)

OWL-Time allows partially-ordered time (allowing, e.g., branching
futures), but one can make it linear with a single assertion.    (09)

John Sowa:
 > a) 4 dimensions or 3+1 dimensions?  A 4-D approach treats objects
 >     and processes as connected regions of a four-dimensional
 >     space-time continuum.  A 3+1 D approach treats space and time
 >     as independent, but related, coordinate systems.    (010)

OWL-Time is a theory of time only, not of space, objects, or
processes.  A theory of each of these things would interact with the
theory of time, and our effort was to make that interaction easy to
accomplish.  The theory of time was subsequently integrated with a
language for describing processes (VERL or Video Event Representation
Language -- google it).    (011)

John Sowa:
 > b) Time points or intervals?  Using real numbers to represent time
 >     coordinates implies that time is divisible into infinitely small
 >     points.  But finite intervals with domain-dependent granularity
 >     are more realistic.    (012)

OWL-Time has both instants and intervals, but if you don't want to use
instants, you don't have to.  For example, the Allen interval
relations are defined in terms of "begins", "ends", and "before" on
instants, but you can take them as primitive if you want.    (013)

The real numbers are intended to be one model of the theory, but other
models are possible too, including finite ones.  The density of time,
if desired, requires one assertion.    (014)

John Sowa:
 > c) Contexts or extra arguments on relations?  If time is represented
 >     by some coordinate system, how are those coordinates associated
 >     with the other representations?  By attaching the time (and/or
 >     space) coordinate to a context box or other delimiter that encloses
 >     the description of a situation at that time?  Or by adding another
 >     argument to every relation to indicate the time when it is true?    (015)

We suggest two possible ways one could interface with the time
ontology.  1.  Include time arguments in all time-dependent
predications.  2.  Reify states and events and make their time a
property of the event with the predicates "atTime(e,t)" for instants,
"during(e,T)" for intervals, and "timeSpanOf(T,e)" for both.    (016)

I don't know how it would be integrated into a "context box" approach
without seeing the syntax, but I can't believe it would be difficult,
unless that logic tied your hands in perverse ways.    (017)

Paola DiMaio:
 > Following on the fact that the ability of representing 'temporal'
 > properties may be desirable
 > in a semantic data model, does your conclusion recommend that IKL is used as
 > representation of K (or data) on the web rather than RDF/OWL, or in
 > combination with rdf/owl
 > or what exactly shall we do?    (018)

If all you care about is expressing temporal facts and not doing
complex temporal reasoning, you can probably stay within OWL, and
that's why we developed the projection of OWL-Time into OWL.  The
predicates you would be using would be well-defined and constrained in
their possible interpretations by the full theory, and any subsequent
effort to extend your system to enable temporal reasoning could then
tap into that.    (019)

John Sowa:
 > IKL is a superset of Common Logic (CL), and those extra
 > features were added to support many systems, such as Cyc,
 > which need that extra expressive capability.  The CLIF
 > and CGIF dialects of CL have been extended to support
 > the IKL features.    (020)

CL is adequate for expressing the full time ontology and for the
theory of processes in VERL, providing one doesn't object to reifying
states and events and some other things (a position which in my youth
I called "ontological promiscuity").  I don't know whether IKL gives
one any processing advantage over that, or whether it rather just has
to do with ontological scruples.    (021)

Matthew West:
 > There is an ISO standard for the representation of time, ISO 8601.
 > I commend its use, rather than coming up with something else.    (022)

OWL-Time has a straightforward mapping of dates and times into the ISO
8601 standard.  But I agree with John Sowa that this is really only a
small part of a theory of time.    (023)

Paola DiMaio:
 > so true. I mean, about the relativity of time, i guess one would have to 
 > is it server time, or user time or greenwiwh, sideral time or what.
 > where is the  'ontology of time' to serve as reference for the  PTim    (024)

OWL-Time provides a way for dates and times to be anchored time zones.
It does not deal with relativistic time, B.C.E., the Islamic calendar,
the Mayan calendar, leap seconds, and other esoterica, but we would
welcome those as extensions.    (025)

Pat Hayes:
 > and since then a fairly 'standard' ontology was put together for the
 > W3C OWL-based SWServices project by Jerry Hobbs and Feng Pan,
 > originally written in a conventional FOL notation and then as far as
 > possible transcribed into OWL.    (026)

Thanks for the plug, Pat.    (027)

Pat Hayes:
 > Do you have any granularity axioms? That is one of the hardest
 > ontological problems, in my experience.    (028)

John Sowa:
 > There are so many hard problems, it's hard to say which are harder.
 > But the idea of taking the least significant digit as the criterion
 > for implicit granularity is fairly common for experimental data
 > (unless some explicit margin of error is stated).    (029)

The basic write-up at    (030)

    http://www.isi.edu/~hobbs/time/owl-time-july04.txt    (031)

has a section on granularity.  It mentions the easy way out that John
refers to, and how that would look in OWL-Time, i.e., silence about
fine-grained properties.  It also discusses how one would go about
making an explicit theory of granularity.  I just reread it, and I
don't _think_ I should be embarassed by it.  It follows my IJCAI-85
paper.    (032)

John Sowa:
 > I would recommend a fairly simple framework for starters, since
 > there's a danger of freezing half-baked ideas before they're fully
 > baked.    (033)

We provide an "entry ontology" for those who don't want to get too
deep.  It's a useful subset of the full ontology.    (034)

Duane Nickull:
 > temperature is another factor which seems to belong in such
 > a framework.    (035)

Time and temperature are two examples of scales.  Not as official as
OWL-Time, but a sketch of a theory of scales can be found at    (036)

    http://www.isi.edu/~hobbs/bgt-scales.text    (037)

Pat Hayes:
 > The result is however rather awkward to read, to put it mildly.    (038)

I beg your pardon!    (039)

Pat Hayes:
 > In practice, I think a single, fairly simple, ontology is best for
 > handling time. It should be agnostic on whether time is ultimately
 > continuous or dense (because either choice gives far more complexity
 > than one ever wants in practice) , ignore branching and circularity,
 > have real but severely limited ability to speak of intermittent
 > intervals like 'every tuesday afternoon', and have a very rich system
 > of times and dates, savvy enough to keep track or real calendars
 > (knows about timezones, summer times, notions of appropriate
 > timescales, etc..    (040)

This is what we tried to do in the entry ontology, although we did not
include a treatment of temporal aggregates like 'every tuesday
afternoon'.  That's in the full theory.    (041)

-- Jerry    (042)

Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/  
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/ 
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (043)

<Prev in Thread] Current Thread [Next in Thread>