Re: RDF* and conjectures

Hi Fabio

I haven’t been following this discussion too closely - but in response to what you said:

> In addition, the whole justification of conjecture is NOT to make unasserted graphs, but to have within the same dataset BOTH unasserted and asserted graphs, and to be able to distinguish which is which. Whatever semantics ends up being chosen of the seven proposed in [1], there will be no semantics for the opposite case, and I would like to be able to express both. 

> You cannot have it both ways: with standard named graphs, either all of them are not asserted, or they all are. 

Tying in PA’s statement:

> The good thing about it is that you are free to define your own semantics and implement it (from scratch, or based on some implementation whose behaviour is compatible with it). That's what nanopublications are doing.


Unsure whether you are aware, but we have done this exercise for N3 [1]: quoted graphs in N3 have very limited meaning, which makes it possible to define any high-level semantics for them - within the same document - by having their triple indicating their semantics, using predicates such as “sem:combinedWithDefault”, “sem:isolatedGraph”, and “sem:onlineRelationTo”, each of which are implemented by N3 rules.


[1] Doerthe Arndt; William Van Woensel. Towards Supporting Multiple Semantics of Named Graphs Using N3 Rules. In: Proceedings of the 13th RuleML+RR 2019 Doctoral Consortium and Rule Challenge, September 16-19, 2019 - Bolzano, Italy, September 16-24, 2019, CEUR-WS.org, 2019.



William


> On Sep 17, 2021, at 5:43 AM, Fabio Vitali <fabio.vitali@unibo.it> wrote:
> 
> Dear Pierre-Antoine, 
> 
> thank you first of all for all the time and energy you spent on our proposal. 
> 
>> From what you wrote, it seems to stem from the misconception that named graphs are asserted, which they are not.
> 
> Thank you. This bears providing some explanations, which I naively skipped over in my first document. 
> 
> The reading of [1] brings us to the conclusion that we DO NOT KNOW if named graphs are asserted. In fact, of the seven different (and reciprocally incompatible) semantics represented in [1], only one (3.4) behaves as if the content of the named graph is not asserted. As far as I can see, all the others do not explicitly rule out the assertion of its content. In addition, it is said: "The RDF Working Group did not define a formal semantics for a multiple graph data model because none of the semantics presented before could obtained consensus. Choosing one or another of the propositions before would have gone against some deployed implementations."
> 
> So there is no clear standing point, wrt plain named graph, that can justify the idea that the content of named graphs is not asserted: depending on your point of view or standing point, they might or might not. 
> 
> In addition, the whole justification of conjecture is NOT to make unasserted graphs, but to have within the same dataset BOTH unasserted and asserted graphs, and to be able to distinguish which is which. Whatever semantics ends up being chosen of the seven proposed in [1], there will be no semantics for the opposite case, and I would like to be able to express both. 
> 
> There are situations in which some graphs should be considered as asserted and other as only postulated. With plain RDF, these two graphs:
> 
> GRAPH :conjecture01 {
>    :Hamlet dc:creator :WilliamShakespeare .
>    :Hamlet dc:created "1603"^^xsd:Year .
> }
> :conjecture01 prov:wasAttributedTo :SamuelJohnson .
> 
> GRAPH :conjecture02 {
>    :Hamlet dc:creator :EdwardDeVere .
>    :Hamlet dc:created "1596"^^xsd:Year .
> }
> :conjecture02 prov:wasAttributedTo :JThomasLooney .
> 
> are either BOTH asserted or BOTH non asserted. Is there a way to distinguish the two situations? Well, in this case we could simply repeat outside of the graph the statements we wish to assert, true. 
> 
> But sometimes (e.g., with nanopublications) I am expected to use graphs. Consider for instance the nanopublication about Ussher's chronology of the Earth that I mentioned in my first message [2]: 
> 
>  :assertion {
>    # Earth inception date is 23rd October 4004 b.C.
>    # Q2 = Earth
>    # P571 = inception date
>    wd:Q2 wdt:P571 "-4004-10-23"^^xsd:date .
>  }
> 
>  :provenance {
>    # Q333481 = James Ussher
>    :assertion prov:wasAttributedTo wd:Q333481 .
>    :assertion prov:hadPrimarySource :AnnalesVeterisTestamenti .
>    :AnnalesVeterisTestamenti a fabio:Work .
>  }
> 
>  :pubInfo {
>    : prov:hadPrimarySource <https://en.wikipedia.org/wiki/Ussher_chronology <https://en.wikipedia.org/wiki/Ussher_chronology>> .
>    : prov:atTime "2021-06-18T15:59:22"^^xsd:dateTime  .
>  }
> 
>  :Head {
>    : a np:Nanopublication .
>    : np:hasAssertion :assertion .
>    : np:hasProvenance :provenance .
>    : np:hasPublicationInfo :pubInfo .
>  }
> 
> The requirements for nanopublications ask me to use four named graphs. Of these four, three are non problematic: :Head, :pubInfo and :provenance simply state objective facts about who, where and how did we gather the information about :assertion. The problem here is with graph :assertion, which contains statements that Ussher thought true, and we should not (namely, that Earth was created on the 23rd of October of 4004 b.C.) 
> 
> You cannot have it both ways: with standard named graphs, either all of them are not asserted, or they all are. 
> 
> In conclusion, whatever semantics you choose for named graphs, it only satisfies one of the two necessary semantics, and yet we need TWO separate constructs for expressing both. These two constructs have not been forwarded, and therefore we must do with what we have. Overall, I believe that it is safer to assume that plain named graphs MIGHT be considered asserted, and therefore propose to introduce a mechanism that clearly and unambiguously MUST NOT be considered asserted. 
> 
> More answers to come on your other message. 
> 
> Thank you again for your time
> 
> 
> Fabio
> 
> --
> 
> [1] https://www.w3.org/TR/rdf11-datasets/ <https://www.w3.org/TR/rdf11-datasets/>
> 
> 
>> On 17 Sep 2021, at 08:18, Pierre-Antoine Champin <pierre-antoine.champin@ercim.eu <mailto:pierre-antoine.champin@ercim.eu>> wrote:
>> 
>> Dear Fabio,
>> 
>> it took me some time to process your email (and I must confess that it still need some processing).
>> 
>> One first remarks, though, which is quite independant from RDF-star (for the moment):
>> 
>> why did you feel the need to "rewrite" the predicates in your conjectures? Why not simply used a named graph? From what you wrote, it seems to stem from the misconception that named graphs are asserted, which they are not.
>> 
>> More precisely, while the RDF specification defines the semantics of a single RDF graph [1], it does *not* defines the semantics of an RDF dataset [2]. In other words, while the semantics of a Turtle file is well defined, the semantics of a TriG file is unspecified. Different alternatives of what the semantics of datasets *could* be have been explored [3]. Some of them (3.3 in particular) make named graph a good fit, I think, for your notion of conjecture, without requiring additional processing or syntax.
>> 
>>  best
>> 
>> [1] https://www.w3.org/TR/rdf11-mt/
>> [2] https://www.w3.org/TR/2014/REC-rdf11-concepts-20140225/#section-dataset
>> [3] https://www.w3.org/TR/rdf11-datasets/
>> 
>> On 25/08/2021 18:07, Fabio Vitali wrote:
>> ...