[Crm-sig] (no subject)

c.e.s.ore at iln.uio.no c.e.s.ore at iln.uio.no
Wed Jun 10 15:43:05 EEST 2009

Dear all,
I commented perhaps to briefly on the diagram. I question the need for the
additional properties P79A and P80A. They represents a fixation and
oversimplification of the start and ending points of the time spans which
is not justified when using the CIDOC CRM as a model for reasoning about
event chronology. Below I have pasted the introduction of presentation Jon
Holmen and I gave at CAA2009 Williamsburg, USA.


In the CIDOC-CRM there are basically two ways to express chronological
information about events. They can be dated relatively to each other:
before, after, overlap etc (CRM properties P114, P117-120).  The events
are seen as wholes with a clear cut beginning and end.

The events can also be connected to a timeline through time-span(s), see
fig. attached.  An E51 Time-Span represents a temporal extension on the
abstraction level, while an E61 Time Primitive is on the implementation
level and is typically represented as an interval on a timeline. An event
has in principle only one time-span. In case an event has more than one
time-span, this expresses diverge opinions about the temporal extension of
the event. Conversely two events may have a common time-span. This
expresses the fact that the events occurred simultaneously.

In cultural heritage databases, it is common to store time determinations
as text, e.g. “the summer 1349” or “active 1450–1455”.  Such textual
defined dates should be mapped to E51 Time-Span in the CIDOC-CRM.  To
enable queries a widespread solution is to represent such text dates as
interval with a numerical start date and an end date. That is, the summer
1349 can be given the interpretation 15-04-1349 to 15-10-1349. This can of
course be extended to all dates. The year 1349 is interpreted as the
interval 01-01-1349 to 31-01-1349. The degree of accuracy can be extended
ad libitum: The day 14-07-1789 is represented as 0000-14-07-1789 to
2400-14-07-1789.  The interpretation of dates as  intervals corresponds to
the E61 Time Primitive in the CIDOC-CRM.  The interpretation of dates as
intervals or time-spans is a powerful model and is easy to implement.

The CIDOC-CRM has a dual view on events. On one side they are viewed as 
entities that can be put in relation to each other by the time operators
P114 is equal in time to, P115 finishes, P116 starts, P117 occurs during,
P118 overlaps in time with, P119 meets in time with, P120 occurs before
(see Allen1984, and Crofts 2009).  When using these operators an event A
can be thought to have an associated time-span with clear cut start point
As and end point Ae.
On the other side, in scientific work a time-span represents an abstract
approximation to the temporal extent of an event. The exact start and end
points are usually not known. In the CIDOC-CRM this approximation to the
temporal extension of events is modeled as two intervals on the timeline,
one for the outer bounds (P81 at sometime within) and one for the inner
bounds (P82 ongoing throughout), see fig. attached. In an ideal situation
with absolute accuracy, these two intervals should be identical. In most
actual research they are not. In many cases there is only information
about an outer interval (at sometime within), that is, by some terminus
post quem (TPQ) and terminus ante quem (TAQ) for the event.

The properties, P81 at sometime within, P82 ongoing throughout open for a
interpretation with fuzzy start and end points of time-spans. The start
points of the two intervals define a third interval (Ass, Ase) in which
the event must have started. This interval can be interpreted as an
approximation to the unknown exact start point of the event. 
Correspondingly, the end points of the inner and outer intervals define a
fourth interval (Aes, Aee) approximating the exact end point of time span
the event, see the attached figure.

The four values or points on the timeline can be used to implement a
reasoning system for time-spans as well as deduction rules corresponding
to the so called Allen operators modeled as P114 to P120 in the CIDOC-CRM.

The objective of any deduction system is to increase the information by
applying rules of interference. In our system the objective is to deduce
the possible relative chronology of events as well as deducing the most
accurate approximation as possible to the actual dates for the events.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: diagram.jpg
Type: image/pjpeg
Size: 78159 bytes
Desc: not available
Url : http://lists.ics.forth.gr/pipermail/crm-sig/attachments/20090610/fd02b7dd/attachment-0001.bin 

More information about the Crm-sig mailing list