Dear Francesco,
Here a more complete comment. I widely agree anyway.
On 3/21/2019 12:49 AM, Francesco Beretta wrote:
Dear Martin, all,
Applying the principles we generally use in conceptual modelling, a
model of « CRM top hierarchy and space&time » like the one you’ll find
in attachment seems plausible. I’ll comment the slides from the top to
the bottom, for the sake of clarity.
Let’s start from the basic modelling principles Martin expressed on
9/3/2019 :
« In the first place, E2 has a substance of "phenomena" something
"becoming" "changing" "moving", "interacting". In addition, we
interpret it now also more statically as including a sort of
maintaining something. It is _necessarily connected_ to some "things"
on which such interactions, changes or temporary, non-essential
formation of properties happen, but we have seen so far no good
general way to describe the ways of involvement at the level of E2.
E92 is nothing of that kind. It is just spacetime, the generalized
space in which we live and think, not what is there not what happens
there. It is just a "where". It is further a volume in that space,
i.e., it must have some inner part, and a surface as fuzzy as it may
be, and a way to identify it. »
As you can see in the attached slides (the more relevant being the
first one), E77 Persistent Item and E2 Temporal entity are
‘phenomenal’ classes. In contrast, Place, Time-Span, STV, Dimension
are ‘regions’ in a reference ‘space’, be this spatial, temporal or
quantitative.
There is a curious difference of the STV to Place, TimeSpan and
Dimensions. Points in spacetime are absolute in the world in which we
live, at least in modern physics, and in our intuition. They are the
"where this happened". If we describe them by coordinates, we use
different reference 'spaces' in the physical space, and in time. But
their identity does not depend on them. Therefore, descriptions in one
reference 'space' can be transformed into descriptions of the same
points in another, if the relations are known. This makes a fundamental
difference.
As we know, E77 Persistent Item instances are in principle not
directly related to time (using properties) but they live in time : we
model this using temporal entities related to these Persistent item
instances « on which such interactions, changes or temporary,
non-essential formation of properties happen » (Martin). Temporal
entities are phenomena we usually perceive or observe in relation to
some persistent item, be this a conceptual or physical one. A very
general property ‘Pxx involves/concern’ would clearly express this
basic phenomenon : persistent items live in time (and space) and we
model this using temporal entities (Events for dynamic moments, Phases
for static characteristics, both phenomenal).
Indeed, but it violates the bottom-up methodology to define a property
which is not well-distinguished from others, and possibly the Open World
assumption, trying to cover all future relations between two classes. It
satisfies the theoretical understanding, but is error-prone for data
modeling. We should only accept if good definitions are found.
E2 Temporal entities (having a phenomenal substance) are projected in
a region in time and these contribute to define their identity (we
stress here the phenomenal aspect, not the epistemological). This
projection in time is modelled as an instance of E52 Time span.
Yes. Without theory or relativity, this projection is unique except for
the zero point. With theory of relativity, it depends on the reference
'space'. With GPS satellites, the correction makes up for an error of
about one km on the surface of Earth, if I remember correctly.
Insofar as we are modelling conceptually we need to keep these two
classes (E2, E52) separate in the model regardless their cardinality
because they are expressions of significantly different substance. The
implementation in an information system, if the (1,1;1,1) cardinality
is chosen, can merge the two classes but this is about implementation
not about the conceptual model : here we must keep both classes
separate for the sake of clarity and consistency with the identity
principles.
I agree.
How do we model, in the next step, projection in physical space ? The
crucial question is: are there any temporal entities without such a
projection ? If yes (e.g. I2 Belief), we have two ways of modelling
this : 1) with a (*1*,1:1,1/n) cardinality for /P7 took place at
/associating it to E4 Period (slide 1) or 2)
Besides that P7 allows for defining wider areas, spacial projections are
not unique. They depend on reference frames, which are in general
moving, be it geologically slow or faster. Therefore (1,1: is wrong).
with a (*0*,1:1,1/n) cardinality associating it directly to E2
Temporal entity (slide 2). This second model would imply in some cases
there is not a projection of a Temporal entity instance in a region in
physical space, although fundamentally there can be one. In the
perspective of simplicity, the second solution would be the preferable
one.
I can only repeat, that if we accept this argument, for questions of
simplicity all properties should be at CRM Entity. It is really a trap,
because the ways an E2 may occur in physical space needs not qualify as
an "extent in space".
But for the sake of consistency with earlier versions of the CRM and
for making the conceptual model more clear and explicit, the first
modelling choice (using E4 Period) would probably be the best one.
This is not a question of the CRM, but a fundamental modelling principle
of ontology. We are committed to understand the "οντα". So far, it has
saved us from many modelling errors and pseudo-solutions, causing
incompatibilities with other "simplified" high-level theories.
Insofar as it has a projection in time and space, an instance of E4
Period (which is composed by Events and Phases) is associated to a
Spacetime Volume. As phenomenon, an instance of E4 Period makes a STV
to be virtually present : if we want to make this explicit, and
especially if we want to explicitly associate a time span with a
place, providing this association with a specific identity,
independently from any E4 Period instance, we need a STV instance (and
the E92 STV class). For this we would then need to have a property
/Pxx has //spacetime//volume/ (1,1:1,1/n) modelled similarly to P4/P7.
In contrast to Place and Time-Span, this volume is unique. It is
(1,1:1,1). It has exactly the same projections to Place and Time as the
E4, and therefore is definitely not similar to P4/P7. Please refer to
texts about CRMgeo.
Incidentally, for all these properties we have to decide if the
maximum cardinality on the side of a Exx Region subclass has to be 1
or n. If we choose ‘n’ we provide a specific identity to the Exx
Region instance, independently from the identity of the related
Phenomenal Class instance. E.g. different E4 Period instances could be
located in the same spatial region, i.e. E53 Place.
Yes, this is correct for P7, but not for the projection. The projection
is unique per reference system. No other phenomenon will have exactly
the same extent. See CRMgeo for more detailed argumentation.
If we think an autonomous identity of E92 STV is not given, and time
and space, and virtual spacetime volume (= time + space) are always
related to at least one temporal entity or period (i.e. to a phenomenon),
This is not the case, because we have the declarative STVs in cRMgeo.
then we could deprecate the E92 Spacetime Volume class and use E4
Period instead as common point of meeting of time and space, in the
phenomenal sense. A E53 Place being the ‘surface’ of a phenomenon
during a given timespan. E93 Presence (as subclass of S4 Observation?)
would be an intersection of time and space in the epistemological
sense, providing an arbitrarily defined snapshot of a Period. As such,
it would have a specific identity and would be modelled as a distinct
class.
Correct. We could move the whole stuff to CRMgeo, and allow it to have a
huge superclass out of CRMbase. That was what we wanted to avoid, but we
could change that.
This way of modelling seems to be more robust and consistent with the
domain : we model phenomena in cultural life associated to persistent
items, phenomena having a projection in time and space, not
time/space/STV as such, independently from phenomena.
If we make multiple ISA, it *means *that both concepts, E2 *AND* E92 are
specialized! The specialization of E92 in E4 is to be in addition
phenomenal. See CRMgeo.
Also, this would avoid the issue of the redundancy of properties which
was the starting point of this discussion : P4/P160 ; P7/P161. For
these reasons I would advocate to abandon the E92 class, knowing that
it is virtually present in E4 Period as its implicit spatio-temporal
surface.
See above. The redundancy of property here is more crying out for a
common superclass on top of E2 and E92.
The last issue I see is the one related to modelling of a STV for a
E18 Physical thing. The /difference/ between the volume occupied by a
physical body as such, be it moving or not, and the volume occupied by
the body moving from one place to another, is not clearly defined if
you make a E18 Physical thing a subclass of E92 Spacetime volume.In
the well known case of the vessels’ fight in Trafalgar, the TSV of the
fight can be treated as projection in time and space of the whole
event, using P4/P7 and, if needed, the correspondent declarative
properties : we have a phenomenon, the fight (E5 as temporal entity),
and a document driven approximation of its STV using in SP10/SP2
(without necessarily the need of using a E92 Spacetime volume/SP7 STV
classes as separated entities with an own identity).
Basically, this is a misunderstanding of the nature of an STV. The STV
is not the result of a backprojection of a space and a time.
Just the opposite. The moving around of the spatial volume of a physical
body IS the STV. It is all spatial points where the volume was one
instant of time after another. he TSTV of the fight cannot be treated as
projection in time and space of the whole event. It is the STV of the
event that has such projections. The resulting box of the two
projections is containing/approximating the event's STV. We need to
understand that!
On the one side, Nelson’s ship itself, as an instance of E18 Physical
thing, would have a time related volume (STV), even being moored in a
port without moving, and the wreck of it on the seafloor has a STV
different from the navigating ship. But this is not about the position
in the fight but about the volume of the Persistent item instance
itself (this volume being a E53 Place instance, as specific « extent
in space » : the 3D ‘surface’ of the ship) and can be modelled using a
Exx Volume/Surface class, modelled as subclass of E3 Condition state.
Well, this is just inventing a new term for the phenomenal STV, if I am
not completely mistaken?
E3 would be modelled directly as subclass of E4 Period or as subclass
of the new Exx Phase class, expressing « phases during the existence
and evolution of an instance of E18 Physical Thing characterized by a
substantial appearance » (Martin). In this case the ‘appearance’ is
the volume of the physical thing understood as a « surface » with a
precise form. The movement of the ship, on the other side, can be
modelled as an instance of E5 Event and associated to the fight E5
Event instance using P9 consists of.
See above. Note, that the spatial projection of the STV of the ship on
the ship as reference space stays within the ship all time, but the
projection of the STV on the seafloor is a long path. The spatial
projection of Nelson's body on the seafloor is within this path., and
within the ship in the other projection.
This approach would allow to have a concise and straightforward model
and avoid inconsistency with the CRM well established way of defining
identity criteria, which isn’t the case if E4 Period and E18 Physical
thing are modelled as subclasses of E92 Spacetime Volume (insofar as
this is not a phenomenon but a ‘region’ in space and time).
Just let me repeat, the inconsistency we are discussing is that the
projection of the event points in an E4 on time occurs as P4 and as
P160, regardless whether we make E4 an STV or not, or if we rename E92
to something else.
All the best,
Martin
All the best
Francesco
--
------------------------------------
Dr. Martin Doerr
Honorary Head of the
Center for Cultural Informatics
Information Systems Laboratory
Institute of Computer Science
Foundation for Research and Technology - Hellas (FORTH)
N.Plastira 100, Vassilika Vouton,
GR70013 Heraklion,Crete,Greece
Vox:+30(2810)391625
Email: [email protected]
Web-site: http://www.ics.forth.gr/isl
