Issue 309: Temporal Primitives

Starting Date: 
2016-08-02
Working Group: 
3
Status: 
On going
Background: 

This is a continuation of the issue 195. For better management of the  discussion about primitives of temporal relations, the issue 195 that was monitoring the discussion about the super properties of  P134 continued (was continued by),   P9 consists of (forms part of), P10 falls within (contains) was closed. The current issue -309- holds the discussion about temporal primitives.

Chryssoula, August 2016


In the 32nd joined meeting of the CIDOC CRM SIG and ISO/TC46/SC4/WG9 and the 25th FRBR - CIDOC CRM Harmonization meeting, it is decided a proposal to be made about a set of temporal relation primitives which are based on fuzzy temporal relations. Also it is decided to be introduced the notion of  temporal distance. MD and Manos Papadakis will elaborate this issue.

Oxford, February 2015

 
Current Proposal: 

Posted by Manos Papadakis and Martin Doerr on 14/5/2015

I am sending you the text about temporal relation primitives which are based on fuzzy temporal relations under the frame of reviewing the scope note of  E7_Activity  P134_continued-by  E7_Activity

Posted by Manos Papadakis and Martin Doerr on 15/5/2015

Please find a revised docx version of temporal relations

In the 35th joined meeting of the CIDOC CRM SIG and   28th FRBR - CIDOC CRM Harmonization meeting the crm-sig decided (a)  the proposed temporal primitives: P173 starts before the end of (ends after the start of), P174 starts before (starts after the start of), P175 starts within (includes the start of), P176  ends before (starts after the end of), P177   ends within (includes the end of), P178 ends after or with (ends before or at the end of) to be introduced in CIDOC CRM (b)   To keep both Allen and these (c)   To add another paragraph in introduction about temporal relations (as compact as possible and including the schematic representations/graphs) which gives the order of these properties and how they relate. (d)    Allen relations to be described as sub properties of these properties (e) editorial work is needed 

Prato, February 2016

posted by Christian Emil on 30/7/2016

Dear all,

I have gone through the paper and also the model as such. M main conclusion is that we sould not just add the end/start point properties as properties between E2 Temporal Entity and that I do not agree with the defintions of the fuzzy operators. I am in the process of finalizing a longer comment on this.

Summary
1)
Thus there is no need for using a common instance of E52 Time-Span to express simultaneousness. So the cardinality of P4 can be restricted to (1,1:1,1).  This will make the model more consistent and P4/P160 will have the function to project a temporal entity/spacetime volume to its temporal component

2)  E52 Time-Span is the pendant to E53 Place. All reasoning about extention in time should be between E52 Time-Span and not between E2 Period. The new properties should be between E52 Time-Span

3) The new properties:
I am not convince that we need the endpoint in
P ΧΧΧ ends within (includes the end of)

Fuzzyness should be defined on the level of time primitives. It is more implementation than ontology. I also suggest we use the method Jon Holmen and I described in the CAA 2009 paper, page 5 left column (attached). We used the already existing properties

P81 ongoing throughout: E61 Time Primitive
P82 at some time within: E61 Time Primitive

which can be seen as a definition of fuzziness of  time spans.

4) There is also an isue about dependencies between

E92 Spacetime volume P161 has spatial projection E53 Place
P4 Period P7 took place at (witnessed): E53 Place

since E92 defines a necessary instance of E53 Place for every spacetime volume and also a neseccary and unique spacetime volume for every instance of E53 Place.

posted by Christian Emil on 31/7/2016

Dear all,

Please find attached a text on the time primitives. The 5 + 1 new properties for time primitives are fine. As I mentioned yesterday there are two issues: Should the domain and range be E52 Timespan? Fuzziness is an implementation issue.

The conclusion in the attached document is:

In the paper from CAA2009 (Holmen & Ore) we argue for possibly overlapping fuzzy boundaries from a deduction point of view. The Allan based algebra is unchanged. If we restrict the relations to non-overlapping fuzzy boundaries, the implementation will still be a model for the Allan-algebra.

The new time primitives may be the basis for an algebra on time intervals. For example, it is possible to express the Allan operators as pairs of the new primitive time operators. Thus a model satisfying the new operators will also be a model for the Allan-algebra. I have not checked in detail, but I assume that we may implement the new primitive with or without possibly overlapping fuzzy borders or without fuzziness altogether and all the implmentations will be models for the algebra. Therefore my view is to leave the fuzziness to the implementation of E61 Time Primitive. I also suggest letting the domain and range of the new time primitives properties to be E52 Time-Span.

posted by Martin on 31/7/2016

On 30/7/2016 10:41 πμ, Christian-Emil Smith Ore wrote:
> Dear all,
>
> I have gone through the paper and also the model as such. M main conclusion is that we sould not just add the end/start point properties as properties between E2 Temporal Entity and that I do not agree with the defintions of the fuzzy operators. I am in the process of finalizing a longer comment on this.
>
> Summary
> 1)
> Thus there is no need for using a common instance of E52 Time-Span to express simultaneousness. So the cardinality of P4 can be restricted to (1,1:1,1).  This will make the model more consistent and P4/P160 will have the function to project a temporal entity/spacetime volume to its temporal component
This makes very much sense to me. The question is, if we can get rid of the intermediate TimeSpan instance.
>
> 2)  E52 Time-Span is the pendant to E53 Place. All reasoning about extention in time should be between E52 Time-Span and not between E2 Period. The new properties should be between E52 Time-Span
>
> 3) The new properties:
> I am not convince that we need the endpoint in
> P ΧΧΧ ends within (includes the end of)
>
> Fuzzyness should be defined on the level of time primitives. It is more implementation than ontology. I also suggest we use the method Jon Holmen and I described in the CAA 2009 paper, page 5 left column (attached). We used the already existing properties
>
> P81 ongoing throughout: E61 Time Primitive
> P82 at some time within: E61 Time Primitive
>
> which can be seen as a definition of fuzziness of  time spans.
I agree with the latter, but the question is, if we can imply from realtions in the phenomenal range
temporal topological terms.

How do you solve:

PXX2 : "Superproperty of:              E7 Activity. P134 continued by
(was continued by[L1] ): E7 Activity"

posted by Christian Emil on 31/7/2016

Dear all,

It was an error (may be more of course) at page 3 where I mixed P81 and P82.

Please find attached a corrected text on the time primitives. The 5 + 1 new properties for time primitives are fine. As I mentioned yesterday there are two issues: Should the domain and range be E52 Timespan? Fuzziness is an implementation issue.

The conclusion in the attached document is:

In the paper from CAA2009 (Holmen & Ore) we argue for possibly overlapping fuzzy boundaries from a deduction point of view. The Allan based algebra is unchanged. If we restrict the relations to non-overlapping fuzzy boundaries, the implementation will still be a model for the Allan-algebra.

The new time primitives may be the basis for an algebra on time intervals. For example, it is possible to express the Allan operators as pairs of the new primitive time operators. Thus a model satisfying the new operators will also be a model for the Allan-algebra. I have not checked in detail, but I assume that we may implement the new primitive with or without possibly overlapping fuzzy borders or without fuzziness altogether and all the implmentations will be models for the algebra. Therefore my view is to leave the fuzziness to the implementation of E61 Time Primitive. I also suggest letting the domain and range of the new time primitives properties to be E52 Time-Span.

 

posted by Martin on 31/7/2016

Dear Christian-Emil,
we will discuss your comments in the meeting.

On 31/7/2016 11:59 πμ, Christian-Emil Smith Ore wrote:
> Dear all,
>
> Please find attached a text on the time primitives. The 5 + 1 new properties for time primitives are fine. As I mentioned yesterday there are two issues: Should the domain and range be E52 Timespan?
Obviously we run into a severe contradiction here between STVs, Timespans, Temporal Entities and Things.
We could make all STVs IsA Time-Span, and E2 IsA Time-Span, if we regard that relativitistic ambiguities of
time reference are irrelevant, and that E52 Time-Span is phenomenal. If we regard the Time Primitive as a
"Declarative Time-Span", then E52 is consistently approximated by the latter. A declarative Time-Span would,
projected on the physical time, still inherit he fuzziness of the date zero reference point, which is updated in regular intervals as time measurement precision increases.

Are there instances of E2 which are indeed without space? Or are we confusing plans with reality?
Now, that we associate a physical thing with an STV, its condition state would again be confined in space.
What about geopolitical units? What about the formation of a "will" to do something? Could we consistently associate with a notion of space?
> Fuzziness is an implementation issue.
I'f argue that fuzziness has two aspects: The deduction/observation point of view, with which I fully agree as you pose it, in which time observation is prior, and the substantial point of view, in which knowledge of processes without time information is prior. Are you sure your proposal covers the latter?

>
> The conclusion in the attached document is:
>
> In the paper from CAA2009 (Holmen & Ore) we argue for possibly overlapping fuzzy boundaries from a deduction point of view. The Allan based algebra is unchanged. If we restrict the relations to non-overlapping fuzzy boundaries, the implementation will still be a model for the Allan-algebra.
>
> The new time primitives may be the basis for an algebra on time intervals. For example, it is possible to express the Allan operators as pairs of the new primitive time operators. Thus a model satisfying the new operators will also be a model for the Allan-algebra. I have not checked in detail, but I assume that we may implement the new primitive with or without possibly overlapping fuzzy borders or without fuzziness altogether and all the implmentations will be models for the algebra. Therefore my view is to leave the fuzziness to the implementation of E61 Time Primitive. I also suggest letting the domain and range of the new time primitives properties to be E52 Time-Span.
The point is that the time primitives can be formulated regarding fuzziness as a generalization of non-fuzziness without affecting the logic and the algebra. This is what we have shown with Manos Papangelis, if we are correct. It further extends into the 4D space without any change. Please contradict!

 

In the 36th joined meeting of the CIDOC CRM SIG and ISO/TC46/SC4/WG9 and the 29th FRBR - CIDOC CRM Harmonization meeting, the sig discussed the points that CEO brought into light through emails and reviewed the document about time primitives proposed by MD.   

About the inconsistencies of STV, Periods, Time spans the decisions were:

Proposal:
 make E2 and STV child of E52, this means that every physical thing gets an on going through out, as well as occurs within  for endurance though it makes more sense to say existing ongoing throughout (for p81). It also makes p4 and p161 redundant
 to think about the difference between perdurants and endurants, MD thinks that for perdurants we look at substance of change, for endurance we look at substance of sameness
 getting rid of p4 is not backwards compatible, but we could build a transformation rule (MD)

Decisions: 
 try to make this model?  MD, CEO, Wolfgang, Carlo, GB, Maria Daskalaki
 Oyvind will criticize, so will Wolfgang.


About the time primitives the decisions were

The following comments were coming up during the discussion:

Proposals made about the notation, About graphics, about text, About the property labels,  About the number of properties we introduce in the text of CRM , About the scope note. The overall decision was to keep the issue on-going. 

Overall decision: on-going. Goal would be to be able to close by next meeting

–         Lida will check super and sub properties..

–         Text improvement by MD and MariaDaskalaki

–         Scope note template improvement by CEO

–         Carlo volunteers will revise formalization of formal properties of temporal primitives

The working document is here

Heraklion,  2/8/2016

In 29/11/2016 Lida sent the HW assigned to her about the sub and super properties of proposed time primitives. The new revised text is here.

Heraklion 29/11/2016


posted by George  on 26/11/2016

Dear all,

Have you had further thoughts about the particular proposal to make E52 a Superclass of STV and E2?

My general reflection on the issue would be

1) This would surely add an element of simplification to the CRM model in that you would be able to reach time reasoning properties quite easily without creating a complicated path

2) Methodologically in terms of good practice I wonder if it would undermine the event oriented approach of CRM in the sense that you could begin to approximate the STVs of things directly… people will start happily putting dates on objects again ignoring their participation in events

I know there are other deeper technical modelling issues to consider in this question, but in order to open a discussion I start there.

In the 37th joined meeting of the CIDOC CRM SIG and ISO/TC46/SC4/WG9 and the 30th   FRBR - CIDOC CRM Harmonization meeting, the  crm-sig made changes to  the text provided by MD and proposed the following:

  1. change the notation of As and Ae  to Astart and Aend ((MD/Lida)
  2. reverse the direction of the last two properties. (MD/Lida)
  3. add the appropriate super/sub relationships to the Allen properties  (MD/Lida)
  4. add examples  (Francesco Berretta and Wolfgang)
  5. to make a new section in the introduction of the CIDOC CRM text with title “Specific Modelling Constructs” and
    (i) move the already existed paragraph “About Types”  there
    (ii) add the introduction of the text of the Temporal Relation Primitives  there.
    (iii) Enumerate the topological relations i.e. spatial and temporal and write an introduction for them. This is assigned to MD&GH 
  6. consider the   semantic grouping of relations/classes  for providing a digital index to tools and for being used in educational materials. This is assigned as HW to GB. 
  7. open new issue for the discussion about relation between E2 and E52 

Berlin, December 2016

Posted by Lida 19/12/2016

The labels of the new temporal relation primitives currently are:

1.P173 starts before or at the end of

2.P174 starts before the end of

3.P175 starts before or with

4.P176 starts before

5.P182 ends before or at the start of

6.P183 ends before

7.P184 ends with or after       

8.P185 ends after         

I think it is better to maintain a uniform format so I would like to suggest the following labels:

1.      P173 starts before or with the end of (ends with or after the start of)

2.P174 starts before the end of (ends after the start of)

3.      P175 starts before or with the start of (starts with or after the start of)

4.      P176 starts before the start of (starts after the start of)

5.      P182 ends before or with the start of (starts with or after the end of)

6.      P183 ends before the start of (starts after the end of)

7.      P184 ends before or with the end of (ends with or after the end of)       

8.      P185 ends before the end of (ends after the end of)

 

posted by Chryssoula on 19/12/2016

Since this issue is about changing labels, please decide via e-mail vote

ACCEPT [ ]

REQUEST MODIFICATION: [....]

by January 15.       

posted by Lida on 17/1/2017

The new updated version of temporal primitives is here

Posted by Martin on 10/3/2017

Dear Richard,

Your comments have been discussed in the last SIG meeting.

Indeed, the fuzzy start and end of a Temporal Entity corresponds to upper and lower bounds for the properties P82a/P81a ('start of the beginning' and 'end of the beginning') and P81b/P82b ('start of the end' and 'end of the end'), . I.e. P82a < Astart < P81a and P81a < P81b and P81b < Aend < P82b.

The idea being, that once the boundaries are fuzzy, they cannot be determined exactly, but only be approximated.

In general, we cannot describe the begin and end as events or phases in their own right, because that would create an endless recursion. This is consistent with P81a < P81b. In cases such events exist, they should be modeled explicitly. Then, knowledge of end of the beginning event may indeed be after knowledge of the start of the ending event.

You said:
 "I must say I'm not particularly convinced by the description of fuzzy boundaries as a 'gradual transition from not existing to definitely existing'.  The sentence "They do not circumscribe the relation of incomplete or imprecise knowledge to reality." suggests that this is nothing to do with uncertainty over when a Temporal Entity started or finished, but I'm afraid I don't understand the meaning of 'circumscribe' in this context.  It would be helpful if this sentence could be re-worded to make its intent clearer."

Please see the current version as decided in the minutes
http://www.cidoc-crm.org/sites/default/files/minutes%20of%2037th%20meeting%20Berlin%20%28draft%202%29.docx.

The phrase has been changed to "The fuzzy boundaries do not describe the relation of incomplete or imprecise knowledge to reality."

It was decided to spell out all boundaries as Astart and  Aend etc. for better reading.

Thank you for your contribution!

Martin

Posted by Gerald on 17/3/2017

Dear all,
I made the enumeration, could you please check if you think its right like that and complete, before we start writing an introduction. I did not include part of relations between E4 Periods or E18 Physical Things, but included relations between E2?
The enumeration is here

Posted by Martin on 24/3/2017

I agree with Gerald's list. We need only the properties that define spatiotemporal, spatial and temporal topological relations. The properties connecting to time/place/spacetime are different I think.

In the 38th joined meeting of the CIDOC CRM SIG and ISO/TC46/SC4/WG9 and the 31st FRBR - CIDOC CRM Harmonization meeting, the crm-sig reviewed

(a) the labels of the temporal properties and decided the following new labels:

P173 starts before or with the end of (ends after or with the start of)
P174 starts before the end of (ends after the start of)
P175 starts before or with the start of (starts after or with the start of)
P176 starts before the start of (starts after the start of)
P182 ends before or with the start of (starts after or with the end of)
P183 ends before the start of (starts after the end of)
P184 ends before or with the end of (ends with or after the end of)
P185 ends before the end of (ends after the end of)
Homework assigned to Lida to add caption to each graphic. Also remains pending the HW about adding examples to the above properties. This HW has been assigned to Francesco Berretta and Wolfgang
 
(b) reviewed the grouping proposed by Gerald and decided the following:
  •  to be removed from the list of spatiotemporal relationships the P166 was a presence of (had presence)  and the  P167 was at(was place of) since they are not primary topological relationships
  •  to accept the groupings proposed by Gerald
  •  to assign to Gerald to write an introduction to these. 
  •  to set an open question remains if these properties are  true super properties to the classes within a model like OGC.

Finally the sig motivated by the open question decided that it is needed  a guideline to be written  about the case of  someone using an extension   would like to reduce to core CRM model. The sig assigned to CEO to write this guideline and  assigned to CEO and Gerald to make the appropriate adjustments  for all CRM extensions. In addition the sig decided to open a new issue for these guidelines.

Heraklion, April 2017