Issue 248: Upgrade P5 consists of and P9 consists of

Starting Date: 
2014-04-25
Working Group: 
4
Status: 
Done
Closing Date: 
2015-02-10
Background: 

Posted by Athina on 12/5/2014 
Dear all, 

About the properties P5 consists of and P9 consists of: is there a conceptual distinction between these properties? The one is about the decomposition of a condition state and the other is about the decomposition of a period, but both classes are classified under same nature, under E2 Temporal Entity; so, what I mean is, why not moving the property “consists of” to the higher class, to E2 Temporal Entity (their superclass), in order to have one property to describe the same thing for both E3 and E4? 

 

Posted by Maliheh Farrokhnia 12/5/2014 

I have another question regarding these properties. We have 3 properties P5, P9, and P45 with the same label of "consists of" or 2 properties P37 and P42 with the same label "assigned". 
I 'm wondering if we can't change the labels of such properties somehow in order not to have exactly the same labels. 

 

Posted by Stephen 12/5/2014 

The question is not could we generalise the property to E2 but are there potential instances of E2 that are not E3's or E4's that potentially do not have decomposition. I do not know and additionally I am not sure I want to spend a lot of time making sure that by their very nature all E2's are decomposable!! If there is a pressing use case for this generalisation then let us, by all means, think about it but ....... 

 

Posted by Simon Spero on 13/5/2014 

>On Mon, May 12, 2014 at 3:53 PM, Stephen Stead 

wrote:

>The question is not could we generalise the property to E2 but are there potential instances of E2 that are not E3's or E4's that potentially do not have decomposition. I do not know and additionally I am not sure I want to spend a lot of time making sure that by their very nature all E2's are decomposable!!

 

This actually a rather significant ontological decision. 
If there are temporal entities that cannot be so divided then the underlying temporal ontology is discrete. 
If every temporal entity can always be so decomposed, then the underlying temporal ontology is dense. 
CRM is committed to a dense ontology (because of the approximate model of time points, and the rejection of any momentary events*) , so it would seem all E2 must be decomposable. 

It is of course, not the case that the type of every part is the same as the type of the whole; conversely, there may be certain granularities where each part is of the same type - e.g. the granularity of a step, each part of a walk is also a walk. 

Simon 
* e.g. "the upward velocity of the ball I just tossed becoming zero" is not considered to be momentary, in spite of calculus, because the precise beginning and end points are cannot be defined as equal, just not distinguishable. 

Posted by Martin 13/5/2014 
Dear Simon, 

I have the impression that you take the density of time, to which CRM commits, for the decomposition of phenomena happening in time, which is E2, the nature of E2 is not the nature of time however. 

Further, you seem to talk about the question if there are minimal elements, that cannot be subdivided further. For time, the CRM does not assume that, as you argue. Currently, the CRM assumes only for Actor minimal elements of decomposition. 

The first question is rather, if there are phenomena in time that reveal a structure with distinct recognizable identities that have a part-whole relationship, such as "Early Minoan" and "Middle Minoan". They are identified by observable characteristics, not by subdivision on the time-line. 

The second question is, if these part-whole relationships can be mixed between subcategories: Can a natural part of a Condition State be an Event, or can a E4 Period ever have a part which is a Condition State? If not, then a common property would produce "non-intended models" as Nicola Guarino describes it. This cannot always be avoided, but we try to minimalize this effect. 

Finally, CRM properties are optional, so there is no commitment given by the definition of a part-of property, that all instances of that class must have parts. Indeed, for material objects, we may end up with nuclear particles as end points, but those are out of scope, but long before, the notion of a Physical Object would loose its meaning. 

If there is a characteristic class in scope which forms the minimal elements, then we would model the decomposition down to this class, as in the case of Actor and Person. 

The question if the labels of these part-of relations should be different is interesting. So far we have prefered that they have all the same name, because that renders the meaning clearly, but different P-number, which renders the constraint. To include the class name in the label may be another method to render the constraint, but it produces long names difficult for translation, and is counterintuitive when used with subclasses, such as an Event consisting of several Periods, or a Period consisting of Events, which is intended. 

Comments? 

Posted by Martin Scholz on 13/5/2014 
the discussion so far is only concerned about part-whole in time. E3 and E4, however, both also extend in space -- but they use different CRM classes: 
an E3 Condition State is limited to an E18 Physical Thing, an E4 Period is limited to an E53 Place. E2, however, is only located in time, not in space. 

The scope notes of P5 and P9 do not explicitly address the spatial aspect. The examples are only about temporal parts. 

I could think of spatial parts like 
* period Renaissance has parts Italian/... Renaissance or 
* the condition state of an ensemble being in ruins but with its parts having different states of decay or even being intact. 

So my question is whether P5 and P9 are intended to only cover temporal or also spatial part-whole-relations. 

Posted by Martin on 13/5/2014 
Hi Martin, 

Indeed the intention is to cover parts regardless spatial or temporal order, because in general they cannot be distinguished. Even "mainly temporal parts", such as a period following another, spreads out over space. Consequently, for some time, both coexist at different places. This is why we are now introducing true spacetime volumes into the CRM. They are more realistic. 

If we feel this should be clearer in the scope note and examples, that could resolve the issue? 
@paveprime.com>

<On 13/5/2014 12:14 μμ, Martin Scholz wrote: 
<Hi, 
<the discussion so far is only concerned about part-whole in time. E3 and E4, 
<however, both also extend in space -- but they use different CRM classes: 
<an E3 Condition State is limited to an E18 Physical Thing, 
<an E4 Period is limited to an E53 Place. E2, however, is only located in time, not in space.

Exactly, this is why my opinion is they cannot be mixed.

Outcome: 

The 32nd joined meeting of the CIDOC CRM SIG and ISO/TC46/SC4/WG9 and the 25th FRBR - CIDOC CRM Harmonization meeting

The crm-sig decide that we can infer from P9 that the part falls within the spacetime volume of the whole. A generalization to E2 would not allow to infer that.  No action should be taken and the issue is closed. Also we decided to post two new issues, the followings:

a) ::Put a comment into the introduction of CRM that same label is used if the substance of the property is regarded to be the same, but different constraints apply to their use with a specific class, such as Parts of a period must be periods etc. (to be elaborated by Øyvind)

b) Put at a visible place the instruction that labels have no meaning

The issue is closed

Oxford February 2015