Re: Factoring of entailment regimes (was: Re: Ill-typed vs. inconsistent?)

On Nov 16, 2012, at 6:28 AM, Richard Cyganiak wrote:

> On 15 Nov 2012, at 21:29, Pat Hayes wrote:
>> I like this ecept I'd like to keep the term "simple entailment" for the basic graph-level case, and so have this:
>> 
>> 1. SImple Entailment
>> IRIs
>> blank nodes
>> triples
>> graphs
>> 
>> 2. RDF with literals
>> Datatype map
>> Typed literals
>> rdf:langString
>> 
>> 3. RDFS (including datatypes-as-classes, if you like: I agree might as well merge these.)
> 
> Okay, I guess Simple Entailment should be retained.
> 
> Just to be really clear here: I suppose this means that Simple Entailment would still say:
> 
> * if E is a language-tagged string "aaa"@ttt in V, then I(E) = <aaa, ttt>
> * if E is a literal in V that is not a language-tagged string, then I(E) = IL(E)

Yes. 

> Because that's essentially the translation of the current Simple Entailment conditions for literals to the new design.

Right.  So semantics recognizes four kinds of 'node': IRIs, language-tagged literals, typed literals, and blank nodes IDs. (It refers to bnode scopes in giving the truth conditions for bnodeIDs)

> And then RDF-with-literals Entailment would just add a single condition, something like (assuming we don't change the handling of ill-typed literals):
> 
> * if <aaa,x> is in D then for any literal "sss"^^ddd in V with I(ddd) = x , 
>   if sss is in the lexical space of x then IL("sss"^^ddd) = L2V(x)(sss),
>     otherwise IL("sss"^^ddd) is not in LV

or, ( not in LV // undefined ) if we do change it. Yes.

> So, technically speaking, rdf:langString would be handled in Simple Entailment, because that's the datatype that distinguishes language-tagged strings from other literals, and is the only datatype that isn't handled via the datatype map. 

Yes, I guess you are right. I don't think there are any axioms or rules that depend upon just this case, so the simple entailment lemma will still work. 

> 
> (See here:
> http://dvcs.w3.org/hg/rdf/raw-file/default/rdf-concepts/index.html#section-Graph-Literal
> And the second note here:
> http://dvcs.w3.org/hg/rdf/raw-file/default/rdf-concepts/index.html#section-Datatypes )
> 
> All the other conditions (starting to take the meaning of specific IRIs into account, other than the IRIs in the datatype map) would then be in RDFS-entailment.
> 
> That would work for me.

Sounds like we are in violent agreement :-)  Putting all of this into RDFS-entailment means that entailment regimes no longer correspond exactly to namespaces, however (because of rdf:type not rdfs:type). Some people will, I predict, not like this. We ought to call it out in any case as it will confuse some readers. 

> 
> Background: This should make it possible to take the informative entailment rules out of Semantics, and put those for Simple Entailment and RDF-with-literals Entailment as an appendix into RDF Concepts, and those for RDFS Entailment into RDF Schema, each time with a clear statement that the normative form is the model theory found in RDF Semantics.

Yes, I guess :-)

Pat

> 
> Best,
> Richard
> 

------------------------------------------------------------
IHMC                                     (850)434 8903 or (650)494 3973   
40 South Alcaniz St.           (850)202 4416   office
Pensacola                            (850)202 4440   fax
FL 32502                              (850)291 0667   mobile
phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes

Received on Saturday, 17 November 2012 01:03:32 UTC