Re: Our approach to unasserted assertions is ambiguous and lossy [ was: Re: streamlining the baseline]

> On 1. Jul 2024, at 21:42, Peter F. Patel-Schneider <pfpschneider@gmail.com> wrote:
> 
> On 7/1/24 15:03, Thomas Lörtsch wrote:
>> Am 1. Juli 2024 20:04:32 MESZ schrieb "Peter F. Patel-Schneider" <pfpschneider@gmail.com>:
>>> On 7/1/24 12:18, Thomas Lörtsch wrote:
>>>> 
>>>> 
>>>> Am 1. Juli 2024 16:05:45 MESZ schrieb "Peter F. Patel-Schneider" <pfpschneider@gmail.com>:
>>>>> I do not understand this entire line of argument at all.
>>>> 
>>>> I do not understand yours.
>>>> 
>>>> For starters: which graphs? I didn't use graphs in my examples. RDF named graphs? Out of scope as I learned in response to my Nested Named Graphs proposal last autumn. So your technique of separating realms of validity employed below has no representation in RDF.
>>> 
>>> RDF is nothing without graphs.  In particular, assertion can only be understood with reference to an RDF graph - without a graph there is no notion of assertion.  Named graphs are simply graphs, with their own asserted triples.
>> RDF named graphs have no semantics. I have no idea what kind of graphs your curly btacket syntax refers to. Anyway my examples didn't use graphs for goid reason. By introducing some underspecified notion of graphs you changed they topic, and then declared victory. I think that's not helpful.
> 
> RDF named graphs do indeed have semantics - they are RDF graphs and thus have all their semantics.  An RDF graph is a set, so I used {} notation to indicate a set of triples, as is quite normal.
> 
> And RDF graphs do have a well-defined semantics - see https://www.w3.org/TR/rdf12-semantics/
> 
> What you might be alluding to is that references to RDF named graphs don't have semantics.   But they do have semantics in RDF graphs, just not one that is very useful.

Graphs are simply not on the table. We have to find a solution that doesn’t rely on graphs as a means to demarcate and disambiguate. Otherwise I would just refer to the Nested Named Graphs proposal from last automn [0] and be done with it. That said, I’m glad Antoine commented on how you characterize graphs and their semantics. 

>>>> And what about qualified statements a la Wikipedia? How do you propose to express those with the means of RDF-star? Or is that a different problem in your opinion?
>>> 
>>> RDF-star has no notion of qualified statements.   There are only triples. Embedded triples can be used to implement qualified statements (more or less) but the meaning of this implementation comes from RDF not from anything that is not part of RDF.
>>> 
>>> Embedded triples change nothing about assertion.  It is still the case that a triple is either in an RDF graph - and thus true or asserted in that graph - or not in an RDF graph - and thus not asserted in that graph.
>>> 
>>> That said, a way to implement qualified statements from Wikidata uses a triple whose subject is linked to the embedded triple corresponding to the main snak, whose predicate is the predicate of the qualifying snak, and whose object is the value of the qualifying snak.
>> And then, how is that embedded triple specified? Maybe you should read my mail again.
> 
> Just like any other embedded triple is specified.

Well, it is the whole point of my mail - to which you respond and which you say you don’t understand at all - to discuss that specification. It seems you didn’t take my advice to read it again.

> So if you want to represent the given names of Douglas Adams you would create an RDF graph containing the following triples (using the well-known Wikidata RDF prefixes):
> 
> wd:Q42 wdt:P735 wd:Q463035 .
> _:a rdf:reifies (embedded wd:Q42 wdt:P735 wd:Q463035 .)
> _:a wdt:P1545 1 .
> _:a wdt:P3831 wd:Q3409033 .
> 
> wd:Q42 wdt:P735 wd:Q19688263 .
> _:a rdf:reifies (embedded wd:Q42 wdt:P735 wd:Q19688263 .)
> _:a wdt:P1545 2 .
> 
> Here I am just using (embedded ...) for embedded triples instead of any of they proposed syntaxes.

Reification, assuming you take rdf:reifies to mean reification like it is defined in RDF, doesn’t provide a connection between 

> wd:Q42 wdt:P735 wd:Q463035 .

and _:a in

> _:a rdf:reifies (embedded wd:Q42 wdt:P735 wd:Q463035 .)


The former is a fact, whereas _:a just describes the occurrence of such a fact, wherever, whenever, whatsoever. That at least is the semantics of RDF standard reification which I have to assume we all refer to when we say "reification", and when we use terms like "rdf:reifies". 
B.t.w. as far as can see that shared understanding about what "rdf:reifies" means is not specified anywhere, which is why I suggest that we provide a mapping from whatever terms and syntaxes we define to RDF 1.1. 

The problem is that reification is not what LPGs, Wikidata and many other use cases demand. They demand qualification. In Wikidata the fact is qualified, plain and simple. It is an n-ary relation. There is no wobbly, metaphysical questions inducing, on situation and interpretation dependant "sort-of" connection between fact and annotation, where you define reification by what it is not and Bryan answers that in that case he just doesn’t care. Not only practitioners don’t care, but also logicians and semanticists don’t care, as you yourself just now, in this very email of yours have vividly given prove of. If it serves your argument, graphs are well defined, reification provides a solid bond between fact and annotation, etc. I’m sick and tired of these on-off relationships to semantic detail. We need something solid that non-logicians as well as logicians can use without doubt, or sarcasm.

And now I can only refer you to my mail again (the one you claim to not understand at all), because the question of how to specify the semantics of embedded triples if not as reification is discussed there at some length already.

Best,
Thomas


[0] https://github.com/rat10/nng


>> Best,
>> Thomas
>>> peter
>>> 
>>> 
>>>>> The idea seems to be that there is a fundamental difference between the two uses of rdf:reifies in the two RDF graphs
>>>>> {
>>>>>    _:x rdf:reifies (embedded P) .
>>>>> }
>>>>> and
>>>>> {
>>>>>    _:y rdf:reifies (embedded P) .
>>>>>    P
>>>>> }
>>>>> 
>>>>> So that there is a need for two different relationships in the RDF namespace as in
>>>>> {
>>>>>    _:x rdf:cites (embedded :Moon :madeOf :Cheese );
>>>>>        :reportedBy :Alice .
>>>>> }
>>>>> 
>>>>> {
>>>>>    _:y rdf:asserts (embedded :Moon :madeOf :Cheese ) ;
>>>>>        :claimedBy :Bob .
>>>>>    :Moon :madeOf :Cheese .
>>>>> }
>>>>> 
>>>>> because in one case :Bob is claming that the embedded triple is true but the other case :Alice is merely reporting on the embedded triple without claiming that it is true.
>>>>> 
>>>>> 
>>>>> But there is no reason to assert the embedded triple just because :Bob is making a claim about it, nor is there any reason to use different relationships to the embedded triple just because the reifier is used in different ways.
>>>>> 
>>>>> That is rdf:reifies is adequate for both
>>>>> 
>>>>> {
>>>>>    _:x rdf:reifies (embedded :Moon :madeOf :Cheese );
>>>>>        :reportedBy :Alice .
>>>>> }
>>>>> 
>>>>> {
>>>>>    _:y rdf:reifies (embedded :Moon :madeOf :Cheese ) ;
>>>>>        :claimedBy :Bob .
>>>>>    :Moon :madeOf :Cheese .
>>>>> }
>>>>> 
>>>>> whether or not
>>>>> 
>>>>>    :Moon :madeOf :Cheese .
>>>>> 
>>>>> is also asserted in the same RDF graph or there are other reports or claims about the same triple in that graph.
>>>>> 
>>>>> peter
>>>>> 
>>> 
>

Received on Friday, 5 July 2024 10:20:39 UTC