W3C

– DRAFT –
RDF-Star Semantics TF

06 December 2024

Attendees

Present
AndyS, doerthe, enrico, gkellogg, niklasl, pchampin, Souri, TallTed, thomas, william_vw
Regrets
-
Chair
enrico
Scribe
enrico

Meeting minutes

<TallTed> Note that until RRSAgent's arrival, there has been no log ... so present+, meeting:, agenda:, etc., may have to be repeated. (I can't see IRC earlier than 10 past the hour, when I joined.)

<doerthe> fine with that

<doerthe> so, to me, only the rdf:type is open :)

<doerthe> and the subject position of course...

<niklasl> Occurrences (tokens are a kind of occurrences).

<Zakim> thomas, you wanted to say that I don't oppose triple terms in subject position anymore

<niklasl> I'd say conflations are a general problem. And not only in RDF. :)

<niklasl> The notion of distinct identity is the cause of those problems. (But without that, we only have fuzzy logic.)

<Zakim> pchampin, you wanted to apologize for misreading the PR

<niklasl> Question of definition: The classes are the sets of properties that (at times / may) point to or from the *edges* in the graph (and not the nodes)?

<doerthe> that it looks ugly in the spec ;)

<doerthe> it might not stop there

<doerthe> rdf:StarificationProperty

<niklasl> rdf:TripleTermProperty

<pchampin> thomas, you might be interested in https://www.w3.org/TR/2013/REC-prov-dm-20130430/#term-specialization

<william_vw> thomas do you have a link to the Berlin example :-)

<niklasl> We have not been able to make it immediately clear, but we know that it is problematic?

<thomas> @william_vw :Berlin :likedBy :John ; :in :Summer ; :likedBy :Alice ; :in :Winter .

<Souri> The other possibility is to exclude everything except rdf:reifies.

<william_vw> @thomas thanks

<thomas> @Souri ... but then what about rdfs:states ;-) IMO we need more properties than just rd:reifies

<Souri> or rdf:associates, just to be neutral -- :r rdf:associates <<( :s :p :o )>> . It is just an association and nothing else.

<Souri> @thomas rdfs:states probably carries too much meaning

<thomas> @pachampin but if it has no meaning then it just states the obvious

<Souri> An assembly language (like RDF) can provide just the minimal core support => :r rdf:associates <<( :s :p :o )>> . :r a :Saying . This is as opposed to (in a higher level language): :r :says <<( :s :p :o )>> .

<niklasl> I would be satisfied with that.

<Souri> I like the minimality.

<william_vw> +1

<niklasl> (I did include no use as subject outside of generalized RDF in what I would be satisfied with.)

<doerthe> and that was where we started, right?

doerthe: ofc

<Zakim> thomas, you wanted to propose "rdf:indicates" (which Enrico proposed months ago)

Etymology. The term "reification" originates from the combination of the Latin terms res ("thing") and -fication, a suffix related to facere ("to make").

Reification is when we make a "thing" (a resource) out from something abstract (a triple term)

<thomas> Yes, but in my intuition that somehow seems to imply that the thing exists (or is brought into existance by the reification) but that is not what is happening (at least not with rdf:reifies as opposed to rdfs:states)

<doerthe> but in my opinion it is not that clear, what the triple term denotes

<doerthe> (and it is not even important in my opinion )

<doerthe> "john has four letters" vs. "john loves paul", even without triple term

<Souri> :r1 rdf:reifies <<(:john :marries :paula)>> ; a :Photo . :r2 rdf:reifies <<(:john :marries :paula)>> ; a :Event .

yes, Souri

<doerthe> I think we should keep things open even if we reify, we do not need to know what the triple term denotes and therefore, I would leave things that open that I would not even add injectivity

<Souri> No.

I think we should discuss this injectivity stuff once and for all doerthe

<doerthe> sorry for always bringing it up, the cicero in me ;)

🤺🤺\

<thomas> we shouldn't kill rdf:TripleTerm

<gb> Issue 27 Integrating different ontology designs through entailment upon triple terms (by niklasl) [use case]

<william_vw> @niklasl no pun intended, I'm sure

<Souri> Not sure I understand this discussion about appropriateness of injectivity. Can we take this example to explain this? => :r1 rdf:reifies <<(:john :marries :paula)>> ; a :Photo . :r2 rdf:reifies <<(:john :marries :paula)>> ; a :Event .

<niklasl> This, by the way, is why I'm not really comfortable with the type rdf:Triple*Term*. It's not the type of that resource, but of the syntactic structure?

<Souri> Is this what you are saying? Two entities: [SR: Set of Reifiers], [ST: Set of Triple-Terms]. One possibly many-to-many relationship: rdf:reifies that is a subset of SR x ST?

<william_vw> is this the problem - without injectivity, we can have:

<william_vw> r1 --reifies--> RE1

<william_vw> where RE1 = <<:john :loves :paul>>, <<:mary :loves :paula>>

<william_vw> so, no longer know what r1 is an occurrence of? ...

<niklasl> Yes.

<niklasl> (AFAICS)

<thomas> r1 is then an occurrence of those two triples

<niklasl> Yes. I think ReflexiveProperty came up before as a counter, but the directional aspect is still true (only one implies the other).

<niklasl> SymmetricProperty, sorry.

<Souri> So, with injectivity: <<( :s1 :p1 :o1 )>> owl:sameAs <<( :s2 :p2 :o2 )>> . <= => :s1 owl:sameAs :s2 . :p1 owl:sameAs :p2 . :o1 owl:sameAs :o2 .

<niklasl> Yes.

<william_vw> (the above is also known as a "fan trap" in databases, I believe :-) )

<william_vw> to avoid it, we only need a many-to-one here, no?

<Souri> So, can s1-s2, p1-p2, o1-o2 "sameAs" be inferred from the following triples? (I hope not because there will be issues if rdf:reifies was used in one-to-many manner here.) => :r1 owl:sameAs :r2 . :r1 rdf:reifies <<( :s1 :p1 :o1 )>> . :r2 rdf:reifies << :s2 :p2 :o2 )>> .

<william_vw> @Souri I don't think so, since rdf:reifies is not injective

<TallTed> the more that gets written in IRC, the more I'll know about what is said from here on...

<william_vw> From my end, I think I understand enrico's point. Have to drop off.

<thomas> "it" doesn't refer to a sentence but to what the sentence describes

<thomas> in our semantics the reifier doesn't refer to what the sentence describes, but to the sentence itself

<niklasl> https://en.wikipedia.org/wiki/Tl%C3%B6n,_Uqbar,_Orbis_Tertius

<doerthe> we scared everyone away :D

<thomas> proposition = statement . it can be true or false

<niklasl> +1000 for rdf:Proposition

<thomas> so we're not saying about it's truth value

<thomas> but then we should clarify the distinction to a fact, which we consider a true statement

<niklasl> ... And then.... rdf:PropositionProperty ?

<thomas> "F-Logic is cool, Common Logic not so much"

<thomas> Enrico: "a triple term denotes a proposition"

<thomas> +1 to that

<niklasl> Works for me. Noting that "An IRI denotes a[ny] resource.", and "A literal denotes .[... a value in the value space of the datatype of the literal]"

Minutes manually created (not a transcript), formatted by scribe.perl version 238 (Fri Oct 18 20:51:13 2024 UTC).

Diagnostics

Succeeded: s/just association/just an association/

Succeeded: s/rdf:reifies)/rdf:reifies as opposed to rdfs:states)

Succeeded: s/F-Logic is cool/"F-Logic is cool, Common Logic not so much"

No scribenick or scribe found. Guessed: enrico

All speakers: doerthe

Active on IRC: AndyS, doerthe, enrico, gkellogg, niklasl, pchampin, Souri, TallTed, thomas, william_vw