Defining some transitive object properties

[florenceclavaud]

It would be useful to define new transitive object properties, in addition to the existing ones, whenever it may significantly help query and explore a RiC-O knowledge graph.

In particular, rico:includesOrIncluded could be used for direct relations only (and thus its semantics would be more precise), and a new transitive property could be added for inferring indirect inclusion relations, like when a fonds directly includes a series, the series directly includes a file, and the file includes an item : the new transitive property 'indirectly includes or included' (or something like that) between the fonds and the item, and the fonds and the file, could then be inferred. This would help process RiC-O graphs (as there may be a huge quantity of direct inclusion relations in such graphs; querying the graph through rico:includesOrIncluded+ may then take a lot of time). Such a new indirect, transitive relation could also probably be directly asserted (I mean expressed, not inferred) in some cases, where for example you know that a file is or was part of a fonds, without knowing exactly any more (or yet) in which exact location it was (or is).

rico:hasOrHasSubdivision (the partitive relation between Groups) may be another interesting use case.

Another use case could be the temporal relations, like rico:precedesInTime.

This at least should be investigated systematically (which properties would be concerned, what to do and how exactly, check the consequences).

BTW this idea was suggested to me in 2020 by @tfrancart, when we began to think of our demonstrator project. I put it aside, and now it is time to work on it.

[florenceclavaud]

One more word about this: as the tense used in the verb of some of these existing relations is both present and past, we have to be very careful. Do direct relations of the kind always mean that the indirect ones can be inferred? In the end IMHO it is not certain. For temporal relations, yes of course. For partitive ones, let us think of this more. At least, they could be inferred if the tense is present only.

[williamsonrichard]

As it happens, I was just pondering yesterday whether rico:isOrWasSubordinateTo could be transitive. As you suggest, the semantics are more precise without the transitivity, but the idea to introduce a new, indirect relation sounds like a good one if one can think of a good nomenclature.

Your point about verb tense is a good one. It may be worth comprehensively reviewing whether one should split up isOrWas properties into two, where was properties could for example be required to have a Date in their domain, e.g. isEndDateOf some Date.

If one has both a direct and an indirect relation, it may be useful to make use of composition of properties, i.e. have something like isInHierarchyOf being a sub-property of isSubordinateTo o isSubordinateTo and isSubordinateTo o isInHierarchyOf or similar.

[florenceclavaud]

Yes! thanks @williamsonrichard for your comment. We are going to investigate this and come back with precise proposals.

[jbkrause]

@regineheberlein and I suggest to add transitive properties for the following properties, with the caveat that it would add 25 properties to an already complex ontology

RiC-O_relations_transitivity_2023-04-19.xlsx

[florenceclavaud]

OK, let us for now discuss this complex topic within the EGAD RiC-O subteam. And then we will of course come back here.

[florenceclavaud]

Hi @williamsonrichard , just to let you know that after discussing the topic, we have started to work on this, in a dedicated branch (issue-41-define-transitive-properties).

[williamsonrichard]

Great, thank you!

[florenceclavaud]

Work done, after quite intense discussions, by several persons in the team. Everything merged in the version_1-0 branch. So the corresponding PR (#68) is closed now.

The transitive properties created, with new accompanying properties, are subproperties of:

• followsOrFollows
• hasOrHadPart
• hasOrHadSubordinate and of their inverse properties.

If somebody here wants to use this version_1-0 branch, please remember it is a development, unstable branch, which may contain components that may be at an intermediate stage, thus not usable.