- From: Pierre-Antoine Champin <pierre-antoine.champin@ercim.eu>
- Date: Wed, 15 Dec 2021 10:12:08 +0100
- To: Anthony Moretti <anthony.moretti@gmail.com>
- Cc: "public-rdf-star@w3.org" <public-rdf-star@w3.org>
- Message-ID: <62aa7182-16a2-d787-55a3-2b3e41ee47f3@ercim.eu>
Hi Anthony,
I am not a big fan about "refersTo", and actually I am having second
thoughts about "mentions" as well, at least in the use-case addressed by
example 8 [1] in the CG report. The reason is that "mention" (and, to
some extent, "reference") points to the use-mention distinction [2].
However, the intended meaning of the example was that the triple was
actually *used* in file1.ttl and file2.ttl, not merely mentioned...
I guess this means that we would probably need two distinct
referentially opaque properties: useOf and mentionOf. And here's another
worm out of the can...
pa
[1]
https://w3c.github.io/rdf-star/cg-spec/editors_draft.html#occurrences-example
[2] https://en.wikipedia.org/wiki/Use%E2%80%93mention_distinction
On 15/12/2021 08:18, Anthony Moretti wrote:
> I had one more thought actually, I think I can take it to its logical
> conclusion. Perhaps the pair should be:
>
> occurrenceOf (referentially transparent)
> refersTo (referentially opaque)
>
> The domain of the second would be References, which themselves can be
> referred to, and so on.
>
> References refer to things, and the relevant section of the report
> changes from “Triples and occurrences” to “Triples and references to
> triples”.
>
> In any case the point is to clear the way for the use of
> “occurrenceOf” in a referentially transparent way rather than the way
> it’s used now.
>
> And yes, sorry, side message with Pat.
>
> Regards
> Anthony
>
> On Wed, Dec 15, 2021 at 5:28 PM Pierre-Antoine Champin
> <pierre-antoine.champin@ercim.eu> wrote:
>
>
> On 15/12/2021 05:45, Anthony Moretti wrote:
>>
>> Note that the `:occurenceOf` predicate in the CG report is
>> just an *example*, it has no "official" standing in the CG
>> report. Yes, I avoided that term in order to avoid confusion
>> with the discussion in the report, but I do not consider any
>> term as definitely taken. "occurrence" vs.
>> "literalOccurrence" would be fine by me, I guess.
>>
>> Note however that we have tried, in the RDF-star group, to
>> coin a general vocabulary for occurrences... but we failed to
>> reach consensus after a long discussion, and decided to defer
>> that to the future working group. This is a nasty can of
>> worms, with a lot of subtle distinctions.
>>
>>
>> If I've found the right discussions you're right about it being a
>> can of worms (169 <https://github.com/w3c/rdf-star/issues/169>,
>> 170 <https://github.com/w3c/rdf-star/issues/170>, 209
>> <https://github.com/w3c/rdf-star/pull/209> etc). I read through
>> them and still think that "occurrenceOf" is preferable to
>> "realizationOf". After some thought, and also heeding Pat's
>> advice about avoiding the use of "literal",
> did I miss some message in the thread? Or are you referring to
> another discussion with Pat?
>> maybe the relations of concern could be:
>>
>> occurrenceOf (referentially transparent)
>> mentionOf (referentially opaque)
> I like that.
>>
>> In the dictionary, a "mention" is "a reference to someone or
>> something". Every time I talk about a triple, whether in writing
>> or speech, can be described as a mention of that triple.
>>
>> So my proposal in this thread is something more targeted: a
>> property whose domain is restricted a given Event class
>> (although, as Peter Rivett poinrted out, schema:Event is
>> maybe not the most appropriate one).
>>
>>
>> There has been lots of discussion
>> <https://lists.w3.org/Archives/Public/public-schemaorg/2018Jun/0013.html>
>> (that I've been a part of) where people would like to broaden the
>> scope of schema:Event to include such things as historical events
>> and notable time periods. If schema:Event were to use the
>> dictionary definition in its description ("a thing that happens
>> or takes place, especially one of importance") then it might
>> solve those problems and also be the domain for "occurrenceOf".
> Good to know, thanks for the pointer.
>
>>
>> Regards
>> Anthony
>>
>>
>> On Wed, Dec 15, 2021 at 12:34 AM Pierre-Antoine Champin
>> <pierre-antoine.champin@ercim.eu> wrote:
>>
>>
>> On 14/12/2021 14:41, Anthony Moretti wrote:
>>>
>>> Exactly. What I like about the "realization" pattern (or
>>> any similar pattern involving quoted triples) is that it
>>> keeps a link between the complex construct (the event)
>>> and the simple triple (asserted or not).
>>>
>>> pa
>>>
>>>
>>> It's a good solution that preserves all the information.
>>>
>>> In place of "realizationOf" I'd like to suggest renaming it
>>> to "occurrenceOf". If we're discussing reoccurring
>>> relationships then each Event is an occurrence of that
>>> relationship.
>>>
>>> As a follow-on I'd also suggest renaming theexisting
>>> <https://w3c.github.io/rdf-star/cg-spec/editors_draft.html#occurrences>"occurrenceOf"
>>> relation to "literalOccurrenceOf" (and possibly renaming
>>> that section of the report to "Triples and literal
>>> occurrences of triples").
>>
>> Note that the `:occurenceOf` predicate in the CG report is
>> just an *example*, it has no "official" standing in the CG
>> report. Yes, I avoided that term in order to avoid confusion
>> with the discussion in the report, but I do not consider any
>> term as definitely taken. "occurrence" vs.
>> "literalOccurrence" would be fine by me, I guess.
>>
>> Note however that we have tried, in the RDF-star group, to
>> coin a general vocabulary for occurrences... but we failed to
>> reach consensus after a long discussion, and decided to defer
>> that to the future working group. This is a nasty can of
>> worms, with a lot of subtle distinctions.
>>
>> So my proposal in this thread is something more targeted: a
>> property whose domain is restricted a given Event class
>> (although, as Peter Rivett poinrted out, schema:Event is
>> maybe not the most appropriate one).
>>>
>>> Regards
>>> Anthony
>>>
>>>
>>> On Tue, Dec 14, 2021 at 6:45 PM Pierre-Antoine Champin
>>> <pierre-antoine.champin@ercim.eu> wrote:
>>>
>>>
>>> On 11/12/2021 05:33, Anthony Moretti wrote:
>>>>
>>>> Idea:
>>>>
>>>> define a schema:realizationOf property, whose
>>>> domain is schema:Event and range is rdf-star:Triple
>>>> (with an inverse property schema:realization). The
>>>> above could be expressed in JSON-LD-star [1] as
>>>> follows:
>>>>
>>>> {
>>>> "@context": "https://schema.org/"
>>>> <https://schema.org/>,
>>>> "@type": "Event",
>>>> "realizationOf": { "@id": {
>>>> "@id": "#bowls_club",
>>>> "captain": "#bob"
>>>> }},
>>>> "startDate": "01-01-2019",
>>>> "endDate": "31-12-2019"
>>>> }
>>>>
>>>> That works. Although when stating this I think the
>>>> start and end dates should also be in the RDF-star
>>>> triple. If the dates aren't there then the event is
>>>> adding information to the triple, whereas I think the
>>>> intention of "realization of" is to show a one-to-one
>>>> mapping, is that right?
>>> no, see below
>>>> If the intention isn't a one-to-one mapping then it's
>>>> sort of saying "instance of", where the only thing
>>>> differentiating instances is the time period, which
>>>> implies that all standard RDF triples without start and
>>>> end times are implicit *types* of events (also makes
>>>> sense to me).
>>>
>>> yes, this is the idea behind my examples.
>>>
>>> Note that, by design, RDF-star does not support the
>>> one-to-one mapping, because quoted triples are (roughly)
>>> like IRIs or literals: they represent the *same thing*
>>> everywhere they appear. This is discussed in the CG
>>> report [1].
>>>
>>>>
>>>> [
>>>> a :TenuredOfficeEvent ;
>>>> schema:name "Presidency of the United States"@en ;
>>>> :hasTermStartYear "1885"^^xsd:gYear ;
>>>> :hasTermEndYear "1889"^^xsd:gYear ;
>>>> :hasOfficer dbpedia:Grover_Cleveland
>>>> ] schema:author [
>>>> a schema:Person;
>>>> schema:worksFor <https://www.whitehouse.gov/#this>
>>>> ] ;
>>>> schema:publisher
>>>> <https://www.whitehouse.gov/#this> .
>>>>
>>>> [
>>>> a :TenuredOfficeEvent ;
>>>> schema:name "Presidency of the United States"@en ;
>>>> :hasTermStartYear "1893"^^xsd:gYear ;
>>>> :hasTermEndYear "1897"^^xsd:gYear ;
>>>> :hasOfficer dbpedia:Grover_Cleveland
>>>> ] schema:author [
>>>> a schema:Person;
>>>> schema:worksFor <https://www.whitehouse.gov/#this>
>>>> ] ;
>>>> schema:publisher
>>>> <https://www.whitehouse.gov/#this> .
>>>>
>>>> ## Turtle End ##
>>>>
>>>> Key point:
>>>>
>>>> No reification required, courtesy of RDF's
>>>> fundamental essence :)
>>>>
>>>> Kingsley
>>>>
>>>>
>>>> That works, although it's less flexible because it
>>>> interleaves concepts. For it to be fully understood a
>>>> reasoner has to understand "Presidency of the United
>>>> States" rather than simpler concepts that can be reused
>>>> like "is President of" and "United States". Composition
>>>> over inheritance
>>>> <https://en.wikipedia.org/wiki/Composition_over_inheritance> could
>>>> probably make the design simpler, but yeah it works for
>>>> sure too.
>>>
>>> Exactly. What I like about the "realization" pattern (or
>>> any similar pattern involving quoted triples) is that it
>>> keeps a link between the complex construct (the event)
>>> and the simple triple (asserted or not).
>>>
>>> pa
>>>
>>> [1]
>>> https://w3c.github.io/rdf-star/cg-spec/editors_draft.html#occurrences
>>>
>>>>
>>>> Personally, modeling would be much cleaner and more
>>>> complete if all statements could have start and end
>>>> time positions, and ideally a location position, then
>>>> every statement has the _option_ of being scoped in
>>>> space and time. The modeling of recurring events then
>>>> falls out of that and people could either do it
>>>> Kingsley's way with events or just use statements with
>>>> start and end times, whichever they prefer.
>>>
>>>
>>>>
>>>> Regards
>>>> Anthony
>>>>
>>>>
>>>> On Sat, Dec 11, 2021 at 3:43 AM Pierre-Antoine Champin
>>>> <pierre-antoine.champin@ercim.eu> wrote:
>>>>
>>>>
>>>> On 10/12/2021 04:05, Anthony Moretti wrote:
>>>>> Agreeing with Dan here, you could argue that any
>>>>> instance of schema:Event is also an example.
>>>> +1
>>>>>
>>>>> Taking Simon's example:
>>>>> Bob - is captain of - Bowls Club - Jan 1, 2019–Dec
>>>>> 31, 2019
>>>>> Bob - is captain of - Bowls Club - Jan 1, 2020–Dec
>>>>> 31, 2020
>>>>>
>>>>> Seems equivalent to:
>>>>>
>>>>> schema:Event
>>>>> Bob's captaincy of Bowls Club 2019
>>>>> startTime: Jan 1, 2019
>>>>> endTime: Dec 31, 2019
>>>>>
>>>>> schema:Event
>>>>> Bob's captaincyof Bowls Club 2020
>>>>> startTime: Jan 1, 2020
>>>>> endTime: Dec 31, 2020
>>>>
>>>> Idea:
>>>>
>>>> define a schema:realizationOf property, whose
>>>> domain is schema:Event and range is rdf-star:Triple
>>>> (with an inverse property schema:realization). The
>>>> above could be expressed in JSON-LD-star [1] as
>>>> follows:
>>>>
>>>> {
>>>> "@context": "https://schema.org/"
>>>> <https://schema.org/>,
>>>> "@type": "Event",
>>>> "realizationOf": { "@id": {
>>>> "@id": "#bowls_club",
>>>> "captain": "#bob"
>>>> }},
>>>> "startDate": "01-01-2019",
>>>> "endDate": "31-12-2019"
>>>> }
>>>>
>>>> (assuming that "realization" and "captain" are part
>>>> of the schema.org <http://schema.org> context)
>>>>
>>>> The annotation syntax could also be used, if bob
>>>> was *currently* captain of the club:
>>>>
>>>> {
>>>> "@context": "https://schema.org/"
>>>> <https://schema.org/>,
>>>> "@id": "#bowls_club",
>>>> "captain": {
>>>> "@id": "#bob",
>>>> "@annotation": {
>>>> "realization": {
>>>> "@type": "Event",
>>>> "startDate": "01-01-2021",
>>>> "endDate": "31-12-2021"
>>>> }
>>>> }
>>>> }
>>>> }
>>>>
>>>> pa
>>>>
>>>> [1] https://json-ld.github.io/json-ld-star/
>>>>
>>>>
>>>> PS: in case anyone is wondering, the Turtle-star
>>>> corresponding to the above JSON-LD-star would be
>>>>
>>>> [] a s:Event ;
>>>> s:realizationOf << <#bowls_club> s:captain
>>>> <#bob> >> ;
>>>> s:startDate "01-01-2019"^^s:Date ;
>>>> s:endDate "31-12-2019"^^s:Date.
>>>>
>>>> and
>>>>
>>>> <#bowls_club> s:captain <#bob> {|
>>>> s:realization [
>>>> a s:Event ;
>>>> s:startDate "01-01-2019"^^s:Date ;
>>>> s:endDate "31-12-2019"^^s:Date
>>>> ]
>>>> |}.
>>>>
>>>>>
>>>>> It seems natural to me that every triple should
>>>>> have start and end time positions and possibly
>>>>> also a location position. The above examples seem
>>>>> to me like different ways of saying the same
>>>>> thing, albeit the first has more structure. You
>>>>> could argue that schema:Event is just a
>>>>> convenience type for statements with temporal data.
>>>>>
>>>>> YAGO knowledge base is a good example:
>>>>> https://www.sciencedirect.com/science/article/pii/S0004370212000719
>>>>>
>>>>> Regards
>>>>> Anthony
>>>>>
>>>>>
>>>>> On Fri, Dec 10, 2021 at 8:20 AM Cox, Simon (L&W,
>>>>> Clayton) <Simon.Cox@csiro.au> wrote:
>>>>>
>>>>> Captain of the bowls club is another example.
>>>>>
>>>>> (I was in one of these the other day admiring
>>>>> the wooden honour boards – the same names come
>>>>> up repeatedly but not necessary sequentially.)
>>>>>
>>>>> *From:*Dan Brickley <danbri@google.com>
>>>>> *Sent:* Thursday, 9 December, 2021 22:57
>>>>> *To:* public-rdf-star@w3.org
>>>>> *Subject:* OnAgainOffAgain relations - beyond
>>>>> celeb marriage: Org membership
>>>>>
>>>>> The celebrity re-marriage example is
>>>>> interesting and real, but may look a bit
>>>>> artificial or cornercase. A similarly
>>>>> structured situation is much more common -
>>>>> membership of organizations.
>>>>>
>>>>> For example one organization being a member of
>>>>> another.
>>>>>
>>>>> https://www.wikidata.org/wiki/Q51698517 is the
>>>>> International Fact Checking Network (IFCN). It
>>>>> has a notion of membership grounded in review
>>>>> of members w.r.t. their official principles.
>>>>>
>>>>> Verified signatories are e.g.
>>>>> https://www.wikidata.org/wiki/Q30325238 (Full
>>>>> Fact). There are some organizations such as
>>>>> Snopes
>>>>> (https://www.wikidata.org/wiki/Q2287154) who
>>>>> were once members (verified signatories) but
>>>>> who are not currently.
>>>>>
>>>>> Wikidata uses annotations on a
>>>>> https://www.wikidata.org/wiki/Property:P463
>>>>> edge between IFCN and Snopes to give start/end
>>>>> times (
>>>>>
>>>>> 15 April 2017, 5 June 2019). It also points to
>>>>> evidence/source document.
>>>>>
>>>>> As far as I know Snopes have only been members
>>>>> once, but if they were to rejoin it seems
>>>>> Wikidata could accomodate the task of
>>>>> representing this.
>>>>>
>>>>> Until I learn a better name for it that isn't
>>>>> too grandiose, I am calling these "on again,
>>>>> off again" relationships, in honour of the
>>>>> celebrity marriage/divorce usecase.
>>>>>
>>>>> Dan
>>>>>
>>>>> p.s. another example, not quite notable enough
>>>>> for Wikidata to record:
>>>>>
>>>>> I (https://www.wikidata.org/entity/Q56641640)
>>>>> have twice been a member of
>>>>> https://www.wikidata.org/wiki/Q7552326 (AISB -
>>>>> Society for the Study of Artificial
>>>>> Intelligence and Simulation of Behaviour).
>>>>> But then I have multiple times lived in the
>>>>> U.K., or been in various restaurants; how do
>>>>> we scope RDF-Star's applicability? Which of
>>>>> these are reasonable places it could be used
>>>>> for time-scoped relationships?
>>>>>
Attachments
- application/pgp-keys attachment: OpenPGP public key
Received on Wednesday, 15 December 2021 09:12:15 UTC