Issue 588: Common Policy / Method for Implementing the .1 Properties of Base and Extensions in RDF
Post by George Bruseker (22March 2022)
Dear all, I am wondering if we should have somewhere in the main document where you can find all the .1 properties. Presently, to my knowledge, there is no index of where these properties are. For those that use the .1 properties it is difficult to find them all using the specification document. One cannot search for '.1' as this returns too many results.
What do people think?
Best,
George
Post by George Bruseker (23 March 2022)
Dear all,
I am also wondering about implementing .1 constructs in CRM Archaeo. Although CRM Archaeo declares .1 properties that are actually essential to its implementation, those properties are not (I believe) expressed in the RDF that comes with CRM Archaeo. Is there a hidden PC file for CRM Archaeo or does everyone generate their own? Should it have an official version? This seems important in terms of the property classes being given the same nomenclature etc or interoperability.
Maybe I missed them on the site.
Best,
George
Post by George Bruseker (24 March 2022)
Dear all,
Subsequent to another thread I started here I am proposing that there be a conversation about having a standard policy and method for creating, documenting and making available the .1 properties for base and its extensions in the RDF serialization. At present to my knowledge the PC classes are only available for CRMBase and relative to version 6.2.1. Other extensions however also use .1 constructions and, moreover, CRMbase moves forward and its PC classes should hypothetically move with it. Therefore, I propose we discuss, create and implement a common policy for creating this rdf derivative to support rdf implementations that adopt the .1 constructions.
Best,
George
Post by Pavlos Fafalios (24 March 2022)
Dear George, all,
The new version of PC classes file (for CRMbase version 7.1.1) has been already produced (as discussed in the last meeting) and will be soon available through the crm website (see issue 567).
Thinking of it again:
Since the namespace for the PC classes is the same with that of the base classes/properties, why not including them directly within the base RDFS? (i.e. providing at the end a single rdfs file that also includes the PCs).
Properties of properties are part of the official documentation (of the base model, or of an extension), so why not including the corresponding PC classes in the same RDFS file instead of providing and maintaining a second file? (this applies for both the base model and the extensions)
This is just a suggestion without knowing any discussions, or decisions made, when implementing the PC classes for the first time several years ago (there might be a good reason for having a second file...).
Best regards,
Pavlos
Post by Martin Doerr (24 March 2022)
I think this is a good idea!
Martin
Post by Philippe Michon (05-Apr-2022)
Dear all,
I strongly second this proposal. In my opinion, there is a serious lack of guidance as to how we are supposed to use .1 properties. In my opinion, the management of PCs should not only be considered in order to overcome a technological limit. I believe there is some interesting semantics within PCs that would be worth describing more formally.
If I'm not mistaken, .1s are only very briefly described in the CRMbase specification. I quickly found only five snippets (I didn't include the new proposals about .1 properties in 7.2.1):
In the "Terminology" section under "property": " Properties may themselves have properties that relate to other classes (This feature is used in this model only in order to describe dynamic subtyping of properties)."
in the "About the logical expressions used in the CIDOC CRM": "properties of properties, " .1 properties" are named by ternary predicate symbols; conventionally, we use P14.1 as the ternary predicate symbol corresponding to property P14.1 in the role of. "
in the "Extensions of CIDOC CRM": "Existing properties can be extended, either structurally as subproperties, or in some cases, dynamically, using properties of properties which allow subtyping (see section About Types below)."
in "Specific Modelling Constructs" under "About Types": "Analogous to the function of the P2 has type (is type of) property, some properties in the CIDOC CRM are associated with an additional property. These are numbered in the CIDOC CRM documentation with a '.1' extension. The range of these properties of properties always falls under E55 Type. The purpose of a property of a property is to provide an alternative mechanism to specialize its domain property through the use of property subtypes declared as instances of E55 Type. They do not appear in the property hierarchy list but are included as part of the property declarations and referred to in the class declarations. For example, P62.1 mode of depiction: E55 Type is associated with E24 Physical Man-made Thing. P62 depicts (is depicted by): E1 CRM Entity"
and in "CIDOC CRM Class Declarations": "Properties of properties are provided indented and in parentheses beneath their respective domain property."
My questions and comments:
I believe it would be particularly important to clarify if it is possible to use the .1 property of a superproperty with one of its subproperties. For example, is it allowed to use P3.1_has_note with P190_has_symbolic_content? According to my understanding this is not possible, but I do not believe it is explained anywhere. And if it is indeed possible, the PC extension does not allow this semantics.
Another issue is the fact that "PC0_Typed_CRM_Property" is not part of the CRMbase class hierarchy. I can understand the reasons behind this choice, but again these would need to be defined. In the context of our work at CHIN, we would sometimes have appreciated using a "P2_has_type" on a PC class in order to type it more formally. Is this something the SIG would encourage?
There is also an issue with P67_refers_to and P138_represents. The last is a subproperty of the first and both have a .1 property. In the PC extension, this hierarchy is completely lost, which limits the possibilities of inference. I don't know how to manage this according to CRMpc, but this problem should at least be documented if there is no concrete solution. I don't believe identifying PC138_represents as a subclass of PC67_refers_to would completely resolve this issue. Moreover, this representation would again raise the question concerning the inheritance of .1 properties by subproperties.
In the same vein, the question of the relationship that exists between the structure of PCs and their corresponding CRMbase properties is not explicit. If a user wants to implement P67_refers_to and PC67_refers_to in their knowledge graph, how should they implement the whole thing? Should the ontology support this kind of inference (in the RDF), or should it be handled at the query level?
I also believe that it would be important to standardize the names of the PC entities. For example, why are the first letters of "PC0_Typed_CRM_Property" capitalized while other classes are not (eg. PC3_has_note)?
Now that we formalize the .1 properties, I think it would be more than necessary to have an official scope note for each of them. The vast majority are only very poorly documented in the scope note of their respective property. It is also important to mention that "PC0_Typed_CRM_Property", "P01_has_domain", and "P02_has_range" don't have a scope note at all.
It is also difficult to easily find all the information about .1 properties. Would it be useful to have a dedicated section for them in the introduction?
Best,
Philippe
In the 53rd CIDOC CRM & 46th FRBRoo SIG meeting, PF walked the Sig through the current state of the issue -2 rdfs files generated per published CRMbase model (one for the ontology, one for the Property Classes); plus the following subtopics raised by PM:
- List the .1 properties of the CRM in the specification document
This particular point has been addressed (and resolved) by Issue 571: a list of all the .1 properties will be added in a tabular form under current table 4 (properties table). Links from that table to the .1 property definition in the document.
Proposal: add scope notes for .1 properties –not for v7.1.2 because there is no time before it is submitted to ISO but for v7.2.1 and on it is definitely something to consider
It is also relevant for Issue 556
- Proposal to merge the 2 rdfs files (ontology & pc module); apply this policy in extensions too
Objections to that on the grounds that
- it is trivial to implement the .1 properties in a different file, whereas incorporating them to the ontology file would mean that they need to be used.
- if one wants to extend the PC module for an application, they should not have to extend the core model as well
- there are other ways to implement .1 properties, it doesn’t have to be through reification.
Decision: The relation between the CIDOC CRM and the PC module (rdfs implementation of the .1 properties) should be expressed in a comment in the CRMbase file) to minimize confusion
HW: PF to write the statement in the ontology file.
- Provide scope notes for all PC constructs (using ‘rdfs:comment’)
Proposal: since PCxxx is the same as Pxxx from the point of view of ontology, the scope note of PCxxx could consist of a statement along the lines of “this is the materialization of Pxxx to implement rdf reification. The property means (repeat the scope note of Pxxx)”
HW: PF to prepare a draft statement to be reviewed by the Sig
- Provide implementation guidelines (generate rdf triples using PCs; run SPARQL queries)
Two options available:
- Always use the property (e.g., ‘P14 carried out by’) together with its PC class (‘PC14_carried_out_by’) E.g.:
DATASET EXAMPLE
:painting_sistine_chapel a crm:E7_Activity ;
crm:P14_carried_out_by :Buonaroti .
:Buonaroti a crm:E39_Actor .
:instanceOfPC14 a crm:PC14_carried_out_by ;
crm:P01_has_domain :painting_sistine_chapel ;
crm:P02_has_range :Buonaroti ;
crm:P14.1_in_the_role_of :master_craftsman .
QUERY EXAMPLE: Requesting all activities, together with their actors and the role of each actor
SELECT ?activity ?actor ?role WHERE {
?activity a crm:E7_Activity ;
crm:P14_carried_out_by ?actor .
OPTIONAL {
?x a crm:PC14_carried_out_by ;
crm:P01_has_domain ?activity ;
crm:P02_has_range ?actor ;
crm:P14.1_in_the_role_of ?role } }
- Use the PC class without using the property
DATASET EXAMPLE
:painting_sistine_chapel a crm:E7_Activity .
crm:P14_carried_out_by :Buonaroti .
:Buonaroti a crm:E39_Actor .
:instanceOfPC14 a crm:PC14_carried_out_by ;
crm:P01_has_domain :painting_sistine_chapel ;
crm:P02_has_range :Buonaroti ;
crm:P14.1_in_the_role_of :master_craftsman .
QUERY EXAMPLE: Requesting all activities, together with their actors and the role of each actor
SELECT ?activity ?actor ?role WHERE {
?x a crm:PC14_carried_out_by ;
crm:P01_has_domain ?activity ;
crm:P02_has_range ?actor .
OPTIONAL {
?x crm:P14.1_in_the_role_of ?role } }
// the query makes no use of ‘crm:P14_carried_out_by’
Discussion:
- (b) problematic, in the sense that neither the dataset nor the query make use of P14, so one cannot exploit sub/super property information, make reasoning, etc
- (a) unmanageable, in the sense that it requires users to double the effort. It should fall on the users to query for both P01-PCxxx-P02 and Pxx.
- Union SPARQL query (E39 Actor in the range of P14 displayed along with the E39 Actor in the range of P02 has range) as a 3rd scenario (to be included in the statement)
- Given comments above: resolve this by drafting a recommendation document that lays out the two approaches –namely: if data generation process allows case (a) then go for it, if not, the recommendation is to use case (b) but query for both Pxxx and PCxxx. Also include 3rd scenario in the statement.
HW: PF detailed descriptions (and draft a query for case 3).
The other issues addressed in PMs response are to be discussed at a later stage.
Another point raised was the directionality of PC properties (forward going vs. inverse form).
May 2022
Post by Thanasis Velios (7 July 2022)
Dear all,
I am reading through the discussion on issue 588 (https://cidoc-crm.org/Issue/ID-588-common-policy-method-for-implementing-the-.1-properties-of-base-and-extensions-in-rdf) and looking forward to Pavlos's HW.
I was thinking that we should connect in the graph the property with its reified class. E.g.:
crm:PC14_carried_out_by → crm:P0x_reifies → crm:P14_carried_out_by
This way the two are connected in the graph and some automatic querying could be possible to programme.
With the usual apologies if I have misunderstood the discussion.
All the best,
Thanasis
Post by Pavlos Fafalios (12 September 2022)
Dear all,
Please find my homework for issue 588 in the below link (as well as in the issues' folder):
https://docs.google.com/document/d/1oQRkmMUgyOeDsn3ZbPuQ__VtbigS9DVsHjmOtvx16uo/edit?usp=sharing
Apologies for the delay! Feel free to add your comments or send your feedback!
Best regards,
Pavlos
Post by Mark Fichtner (12 September 2022)
Dear all,
nice work, thanks! I think for RDF this is a valid representation, although I am not very happy to add properties that are not in the cidoc crm directly and that are not part of the language itself (like in this case crm:P03_reifies). As a user/reader of the rdf it is simply hard to understand what is part of the cidoc crm itself and what comes due to "workarounds". Even in as a new ontology/file/addon it mixes cidoc crm and non-cidoc crm things.
Also we have a reification concept (E13 Attribute Assignment), I am not sure if we need even more of these.
I'm looking forward to the discussion!
Best,
Mark Fichtner
Germanisches Nationalmuseum
Post by Pavlos Fafalios (1 December 2022)
Dear all,
Please find my revised homework for issue 588 below:
https://drive.google.com/drive/folders/1b0wW70xo2wjxNlWHYDRl7nr-fzYXTchN...
Feel free to add your comments or send your feedback!
Best regards,
Pavlos
Post by Francesco Beretta (6 December 2022)
Dear Pavlos, all
reconsidering this question of the properties of properties and the proposed solution of the properties-classes remain some doubts and interrogations to me, in particular in relation with the best practices in the field of serialization of conceptual models in RDF.
Metadata about properties as instances, i.e. statement, can be expressed with the standard RDF reification or the new RDF* standard.
What is the best practice in the RDF community to express this kind of properties of properties (and also dates, etc. added to properties in conceptual models) ?
And furthermore: are the CRM properties of properties just 'metadata' or do they carry some additional ontological substance ?
The technical solution of 'PC' does not remove all ambiguity: are they in the end properties or classes? and when we talk about adding, as now, labels and scope notes to the PCs they do becomes classes, don't they? what is then their substance? just to be reified properties?
One could come to think that in fact there is more substance but not totally and adequately expressed, and that should be more carefully analyzed like in the case of P14.1 in the role of: E55 Type or P107.1 kind of member: E55 Type.
Take the example of P3.1 has type: E55 Type — "This property allows differentiation of specific notes, e.g., "construction", "decoration" etc." (thank you Pavlos for the work you've done).
If P3 is not just taken as a CRM replacement of rdfs:comment, shouldn't the so called associated 'note' be modelled as an information object of type 'damage description' (chipped at edge of handle) related to the corresponding human-made object by P129 is about. Or 'chipped at edge of handle' would be a E3 Condition or State and the 'note' its description?
But because P3 has E62 String as a range, which "is not further elaborated upon within the model", it becomes —P3 I mean— quite relevant as it captures the characterization of the item itself, its internal structures, appearance etc.
So, again, are there any best practices in other communities of RDF experts that apply to these types of situations that should be analyzed before further specifying a notion of PC that doesn't seem totally justified, or raising ontological analisis issues, instead of using simple RDF reification?
Best
Francesco
In the 55th joint meeting of the CIDOC CRM and SO/TC46/SC4/WG9; 48th FRBR/LRMoo SIG meeting, the SIG reviewed HW by PF presented on the .1 properties RDFS implementation.
For details, check attached document.
Summary of decisions:
- Re. the relevant section where the scope-note of .1 properties will be placed in the specification document (wrt. their domain property definitions): add a section “Properties” that’s placed right below the scope-note of the property and right above the examples section
- Re. the statement concerning the relation of the PC module to the CIDOC CRM base file: approved, needs to be introduced at the end of the CRMbase file.
- Re. the introductory text and instantiation example to be included in the PC module: add the XML comments in v7.1.1 and subsequent stable versions. The enriched file (e.g., for version 7.1.1) can be found here. How the content is represented (xml comments vs rdfs:comment) can be dealt with in a separate issue.
- Re. the explicit marking of properties that do not form part of the official definition of the CIDOC CRM (but are part of its implementation):
HW to PF to revise the statements, also consider including other properties that are part of the implementation and do not fall under typed properties treatment (like P81a/P81b, P82a/P82b etc.) - Points F through I of the HW have not been reviewed for want of time. Can be discussed through the SIG list and be put to an e-vote.
Belval, December 2022
Post by Pavlos Fafalios (2 May 2023) --in response to Francesco Beretta's post on the list (6 December 2023)
Dear Francesco,
You have raised some interesting points which, I think, need discussion (but after closing this issue :) ), including:
- what was the motivation of introducing property classes
- how can we tackle ambiguity of PCs (classes vs properties, etc etc)
- why not directly using the standard RDF reification vocabulary <https://www.w3.org/TR/rdf-schema/#ch_reificationvocab>, where properties can be attached to statements/triples.
I will include your comments in the working document of Issue 588, with a suggestion to open a new issue for discussing and working on them.
Best regards,
Pavlos
In the 56th joint meeting of the CIDOC CRM and ISO/TC46/SC4/WG9 &49th FRBR/LRMoo SIG, PF walked the SIG through the remainder of the HW that had not been reviewed in the 55th meeting.
Summary of decisions:
- The SIG approved the revised statement wrt the constructs that are part of the rdfs implementation (see statement in the appendix—point E).
- Xml comments to be turned into standardized <rdfs:comment> statements
HW: ETz & PF to implement
- Xml comments to be turned into standardized <rdfs:comment> statements
- The SIG approved the proposal to introduce labels & scope notes for PCs in the PC file (see appendix –point F)
HW: ETz & PF to implement - The SIG approved the proposal to add in the PC file a statement per PC property, of the sort:
<crm:P04_represents rdf:resource= “Pxx_label_of_respective_property”/>
(see appendix –point G AND ALSO see point F, the highlighted statements)
HW: ETz & PF to implement
- (point H).
HW: MD & PF to revisit the implementation guideline for PC properties. - (point I). The SIG approved the proposal by PF to not introduce a PC for the inverse directionality of the corresponding property, on the grounds that it would call for both PCx and PCx-I to instantiate the domain of the corresponding Px.1, which would make querying even more complex.
- (point J). The SIG approved the proposal by PF to start a new issue (640) where to discuss the ontological status of PC properties and the kinds of statements one can make about them.
- (Point K). The SIG did not approve the proposal by GB to declare PC0_Typed_Property a subclass of E1. The topic relates to the new issue discussed in (point J), above.
Crete, May 2023
In the 57th CIDOC CRM & 50th FRBR/LRMoo SIG Meeting, having reviewed and approved the HW by PF (implementation guideline for .1 properties), the SIG resolved to close the issue. For the details of the statement see here.
Issue closed
Marseille, October 2023