Re: on entailments and triple terms

> On 17. Aug 2024, at 19:46, Peter F. Patel-Schneider <pfpschneider@gmail.com> wrote:
> 
> TL;DR:  The RDF-RDFS-SPARQL triad is useful but flawed in theory and in practice.  Probably the best that can be done with triple terms is mirror this flawed triad with base facilities in simple entailment, extended triple term entailment in RDFS, and query patterns in SPARQL that implement most of this extended entailment.
> 
> What should make rdf:type and rdfs:subclassOf so powerful together?  It should be their semantics.  Because :dog rdfs:subclasOf :mammal and :fido rdf:type :dog together RDFS-entail :fido rdf:type :mammal we can say that in theory RDFS is a very simple ontology language.
> 
> What does make rdf:type and rdfs:subclassOf so powerful together?  It's the fact that many users of SPARQL write queries using rdf:type/rdfs:subclassOf*. Because of this it is not too wrong to say that SPARQL supports very simple ontologies.  But it would be far better if instead of requiring users to write more-complex queries SPARQL engines just normally worked in the RDFS entailment regime so that users could just use rdf:type to access RDFS-entailed type information.
> 
> So based on the coerced efforts of users it is possible to consider simple entailment an ontology language and SPARQL a query language for ontologies. But this is at least in part an illusion that falls apart when users do not construct their queries to take into account the RDFS semantics of RDF graphs.  For example, many queries do not take into account rdfs:subpropertyOf when querying non-ontology facts.  This has probably led to underuse of rdfs:subpropertyOf.  Even worse, SPARQL is incapable of uniformly querying ontologies that have subproperties of rdfs:subclassOf.  So, for example, SPARQL cannot uniformly query for class instances in Wikidata, as Wikidata has properties that are subproperties of its version of rdfs:subclassOf.
> 
> What does this mean for triple terms?  Ideally there would be an explosion in the number of SPARQL implementations that can handle RDFS+++ entailments related to triple terms as easily as they can handle simple entailment.  Then it would not matter how the meaning of triple terms is split up between simple entailment, RDF, RDFS, RDFS++, and RDFS+++.
> 
> In the absence of this the best that can be done is to provide base facilities in simple entailment, add vocabulary and define extended entailment in RDFS, and hope that users can and will write SPARQL queries that implement much of the extended entailment on top of simple entailment.

RDF* and SPARQL* are defined to require something that might be called a simple * -entailment: any RDF triple is also an RDF* triple [0]. That captures a plausible and useful intuition. It has been implemented. It was successfull. It was deemed useful for bridging the RDF/LPG divide. This WG was chartered to standardize the approach.

This doesn’t say anything about the general usefulness and feasability of entailment on the semantic web. Addressing general questions concerning entailment is not the task of this WG. The RDF*/SPARQL* approach to entailment of statement terms as regular statements seems to work predictably. It seems like it can be used to support the semantics of the proposed "rdfs:states" in SPARQL-star.

Thomas

> 
> peter
> 

[0] Olaf Hartig: Foundations of RDF* and SPARQL* - An Alternative Approach to 
      Statement-Level Metadata in RDF. In Proceedings of the 11th Alberto Mendelzon 
      International Workshop on Foundations of Data Management (AMW), Montevideo, 
      Uruguay, June 2017
      http://olafhartig.de/files/Hartig_AMW2017_RDFStar.pdf

Received on Monday, 19 August 2024 10:13:23 UTC