Wikidata:Events/Data Modelling Days 2023/SemanticWeb
✨---------------✨---------------✨---------------✨---------------✨---------------
Wikidata Challenges in the semantic web community
AWesterinen
✨---------------✨---------------✨---------------✨---------------✨---------------
👥 Number of participants (including speakers):
40 (at 16:02)
43 (at 16:30)
🖊️ Notes & links
Slides: https://commons.wikimedia.org/wiki/File:Wikidata
_Challenges_in_Semantic_Web_Community.pdf
...
❓ Questions and discussions
Olea: any tool to draw and discover subclassing loops?
AWesterinen: Will think about this and propose something
Jan: Wikidata usage instructions (P2559)
AWesterinen: Maybe incorporate this in EntitySchema/design patterns?
James: Chemical entity was a deliberate modelling choice, to model molecular species such as "methane" as a class of molecules, rather than an instance. Of course we don't want to instantiate a particular methane molecule! So 1M chemical species without an instance is not a bug, it was a modelling choice
Also: having no instances != subclass is not being used. Class items have all sorts of uses, other than containing instances
AWesterinen: But how do you know that the hierarchy is correct without a use case? I have seen many examples of people defining a complex hierarchy that captured amazing but unnecessary detail from the perspective of an "expert".
The use case (for a class) is defined by the links, and also the statements that have it as their object, not just the instances
I will review and would benefit from more discussion with you!
Sky: Time qualifiers are one of the most powerful and underutilized parts of the WD model
Discussion about organizations and W3C ontology: https://www.w3.org/TR/vocab-org/
Frédéric says: There is definitely work to be done in organizational modelling. Several aspects of the W3C organization ontology are not fully represented or entirely missing in Wikidata. https://www.w3.org/TR/vocab-org/
Arthur Smith says:Looks like W3C ontology for orgs uses a org:classification property to specify org types (other than the distinction of "Formal" organizations).
Frédéric says:What I like about it is it makes it easy to represent different parts of an organization.
Arthur Smith says: @Fjjulien - we have a wikiproject for organizations: https://www.wikidata.org/wiki/Wikidata:WikiProject_Organizations - maybe should consider changing the model to more follow the W3C model?
AWesterinen: I will join and try to help!
Frédéric says: @Arthur: I would be in favour of adopting a good external organizational ontology (the W3C one or another one), implementing it in Wikidata, and then relating all existing organizational classes and properties to this ontology.
Peter Patel-Schneider says:@Frederic I would go even farther and merge the existing Wikidata classes into the other ontology.
AWesterinen: You want to be careful to only merge relevant parts since many W3C schemas are designed broadly to address many situations, some of which may not be relevant. Ontology reuse is a big issue that was discussed in Ontology Summit 2014. A link to the Communique/summary is http://ontolog.cim3.net/file/work/OntologySummit2014/OntologySummit2014_Communique/OntologySummit2014_Communique_v1-0-0_20140429-1045.pdf.
Peter: So how can the upper level of the Wikidata ontology be fixed up? (I'm distinguishing this from the top level, which I view as classes like second-order class.)
AWesterinen: I will create a Wiki page discussing this and providing examples/code.
James: @Peter: The brutal truth is the top level doesn't get fixed because 99.999% of queries don't care / don't need it. Do we care?
AWesterinen: I care because I need a good understanding of what to query and how to use it/find semantics.
Frédéric Julien says:Are there authoritative upper ontologies from which we could pick one, and start implementing it in Wikidata?
AWesterinen: I like using schema.org as a start. It is practical and understood by many developers. Most upper ontologies are too philosophical for use and deal with distinguishing existence and knowledge.
Sky Bristol says:I mentioned Basic Formal Ontology earlier along with the Common Core Ontologies.
AWesterinen: I have worked on two different projects that started with Common Core and BFO. There were many arguments about the correct superclass to define, or which class to extend that really made no difference to the specific subclass.
AWesterinen: And BFO does not utilize multiple inheritance, which in turn leads to extremely complex designs to work around this. I find this the main reason for problems using it.
🎯 Key takeaways and outcomes
...
...
☑️ Next steps
Initiate a discussion with the Organization WikiProject on the implementation of an external ontology such as W3C's Organization Ontology.
Done: Listed W3C's Organization Ontology as an external resource on the home page of the WikiProject Organization. -Fjjulien
Done: Discussion topic started on the talk page: https://www.wikidata.org/wiki/Wikidata_talk:WikiProject_Organizations#Implementing_the_W3C_Organization_Ontology_in_Wikidata - Fjjulien
Define means to find loops in the subclassing hierarchy
Continue work to align Wikidata and schema.org
Maybe restarting the GitHub work?
Create guidelines on subclassing vs instantiating, and problems when instances violate class advice (such as positions being instantiated vs referenced)
And better explain/position this in the Semantic community