Re: [OWLAPI-developer] Parsing not-quite-OWL , and Using a namespace vs importing an Ontology (was Re: Load ontology result in 0 axioms)

[Timothy R. <tre...@st...>]

Thanks for a much more useful and complete answer than mine was.

I think though that there is a little more to be said with the valid 
languages. In particular, I don't think that the statement

> These are somewhat ad-hoc; the RDF to OWL mapping spec only covers OWL 
> 2-DL, and the tolerant parsing mode has evolved incrementally. 

is quite right.  I will come back to this.

Here is my understanding of the situation.

There are actually three OWL languages that are relevant: OWL 2 DL, OWL 
2 and OWL 2 Full.  The first two languages are described in the 
introduction to the language structural specification:

      http://www.w3.org/TR/2012/REC-owl2-syntax-20121211/#Introduction


These two languages have a very simple relationship: OWL 2 DL is a 
subset of OWL 2 and they use the same semantics.  (OWL 2 DL is first 
introduced in the third paragraph of the introduction).  The OWL api 
handles OWL 2 and probably many reasoners work with the OWL 2 DL subset.

The OWL 2 Full language is introduced in the document

      http://www.w3.org/TR/2012/REC-owl2-rdf-based-semantics-20121211/


This document is describing the semantics of an extension of RDF, so the 
syntax and the semantics of OWL 2 Full are distinct from the syntactic 
structure and semantics of OWL 2.  Any RDF document can be parsed as OWL 
2 Full.  There is a theorem that describes the relationship between OWL 
2 Full and OWL 2:

      http://www.w3.org/TR/2012/REC-owl2-rdf-based-semantics-20121211/#thm-correspondence


So coming back to the statement that triggered this excursion:

> These are somewhat ad-hoc; the RDF to OWL mapping spec only covers OWL 
> 2-DL, and the tolerant parsing mode has evolved incrementally. 

The RDF to OWL mapping specification does cover some ontologies that are 
OWL 2 but not OWL 2 DL.  In particular, I think that the global 
restrictions on axioms that an OWL 2 DL ontology must satisfy are not 
relevant to the RDF to OWL 2 mapping specification:

                 http://www.w3.org/TR/2012/REC-owl2-syntax-20121211/#Global_Restrictions_on_Axioms_in_OWL_2_DL


However the typing constraints of OWL 2 DL do appear to be relevant:

                 http://www.w3.org/TR/2012/REC-owl2-syntax-20121211/#Typing_Constraints_of_OWL_2_DL


I think that the key to understanding these typing constraints is the 
slightly opaque statement: "These constraints are used for 
disambiguating the types of IRIs when reading ontologies from external 
transfer syntaxes."

-Timothy

On 04/04/2014 05:02 PM, Simon Spero wrote:
> [Not addressed at Tim, hence subject line change)
>
> 1:  Using a namespace is not the same as importing an ontology.
>
> The thesaurus fragment in the RDF/XML  file defines an xml namespace, 
> "skos", which has the value "http://www.w3.org/2004/02/skos/core#".
>
> The IRIs <skos:Concept> and <skos:related> are expanded to references to
> <http://www.w3.org/2004/02/skos/core#Concept> and 
> <http://www.w3.org/2004/02/skos/core#related>.
>
> However, these IRIs are just IRIs;  they do not carry along with them 
> any of the axioms defined in the SKOS  ontology.  The OWL API parser 
> will make guesses as noted below, but some of these guesses are wrong, 
> and must, of necessity, be incomplete.
> For example, OWL API will guess that by default, 
> <http://www.w3.org/2004/02/skos/core#related> is an AnnotationProperty.
>
> The fragment declares that <CoreSubjectThesaurus> is a 
> skos:ConceptScheme, and that <Abbreviations> is a skos:Concept. 
>  However, because there are just IRIs, there is no way to infer that 
> these are different things;  nor can we infer that <Land%20claims> is 
> a skos:Concept.
>
> Declaring an OWL ontology, and adding an OWL imports property will 
> bring in all of the available SKOS axioms [well, at least, the ones 
> that could be expressed in OWL 2-DL].
>  <owl:Ontology rdf:about="http://www.thesaurus.gc.ca/concept/">
>   <imports rdf:resource="http://www.w3.org/2004/02/skos/core"/>
>  </owl:Ontology>
>
> We can now make use of the axioms we weren't able to before.
>
>  For example, if the thesaurus has BT relations, a reasoner will be to 
> tell that if A BT B and B BT C then A BT C ; it will also be able to 
> conclude that B NT A.  [However, it will still not be able to deduce 
> that that C RT A is false, as that cannot be expressed using only OWL 
> 2-DL, so is only noted in a comment].
>
>
>
> 2: Valid RDF vs. Valid OWL-2 DL vs Valid OWL 2 vs  OWLAPI
>
>     I think that the short answer is that this is not an OWL 2
>     ontology.   An OWL 2 compliant parser should reject your ontology
>     as unparsable.
>     In your small sample, notice that there are no declarations of
>     anything as owl entities
>
>     It will parse as an OWL 2 Full ontology but the OWL api only has
>     partial support for that language.
>
>
>
> The original file was not valid RDF, due to an illegal space character 
>  in an IRI.  With the errant space properly escaped, the file can be 
> processed by the OWLAPI in LAX (not strict) mode.
>
> In order to load the ontology, OWLAPI has to make some guesses. These 
> are somewhat ad-hoc; the RDF to OWL mapping spec only covers OWL 2-DL, 
> and the tolerant parsing mode has evolved incrementally.
>
> Properties that aren't declared will usually be turned into Annotation 
> Properties, though a range declaration can generate an object or data 
> property as appropriate;  a class assertion will lead to an assumed 
> class declaration, etc.
>
> The results are usually, but not always valid OWL 2-DL, due to the 
> gradual accumulation of repair heuristics over time - e.g. different 
> orders of triples can lead to different results; I don't think it's 
> possible to score a hat-trick anymore (get data, object, and 
> annotation property declarations inferred for the same property, but I 
> would not bet on it :-)   The heuristics are not clearly documented in 
> any one place, and would be a good candidate  for documenting or 
> refactoring - after version 4.0 :-)
>
> The heuristics used may not be the best choices for OWL 2 (Full), but 
> it's not always clear what the correct mapping ought to be when 
> converting to FSS Axioms.
>
> The API could be noisier / more selective when applying fixes and 
> catching exceptions.
>
>
>
>
> ------------------------------------------------------------------------------
>
>
> _______________________________________________
> Owlapi-developer mailing list
> Owl...@li...
> https://lists.sourceforge.net/lists/listinfo/owlapi-developer