RE: Re: Encoding an incomplete date as xsd:dateTime

Antoine,

Thanks for your explanations.

> >> There's already a slight problem in vCard's treatment of bday; it
> >> includes  xsd:gYear, which is not permitted in OWL2-DL, and it does
> >> not include xsd:string, which is available, and which is required
> >> by the RFC.
> >
> > If I declare the use of xsd:gYear in my ontology, can I use it then?
> > I'm not familiar enough with OWL to answer that question myself, but
> > the way I read ยง9.4 of the OWL syntax [1] I can use any datatype. Can
> > someone more familiar with this topic shed some light on this?
> 
> In OWL 2 DL, you can't declare any term in the xsd:, rdf:, rdfs: and
> owl: namespaces, according to the spec. Even if you could, you would not
> be able to define xsd:gYear because any datatype must be either an
> OWL-compatible datatype (see Section 4.1 of the structural spec), the
> special datatype rdfs:Literal, or a custom datatype that is built from
> unions, intersections, enumerations, complements of, or restrictions of
> already defined datatypes (See Section 5.2 and Section 7). With these
> constructs, you would never be able to define the value space of gYear,
> which is disjoint from all OWL-compatible datatypes.

OK.

> Now, you can still use xsd:gYear if you want because OWL 2 DL processors
> do not have to reject all non-compliant ontologies. Actually, most OWL
> processors would not bother much about gYear. Besides, these
> restrictions are for OWL 2 DL ontologies but the OWL specs also specify
> OWL Full ontologies, which are all valid RDF graphs.
> Depending on what tools you expect to be used on your ontology, the
> restrictions might be irrelevant.

The point is probably that we don't know how our customers will use the data. So far, we have tried to be OWL 2 DL-compatible, but when it comes to temporal data, we simply cannot supply xsd:dateTime from our data (we are sometimes proud when we can supply a xsd:gYear). I guess this simply means that temporal reasoning over cultural heritage datasets was not perceived as a use case when engineering OWL. 

> >> OWL2-DL allows for facets on dataTime  to specify a minimum and
> >> maximum time point value, which can be used in restrictions on
> >> individuals to yield the appropriate models;  however, this
> >> approach is not ideal.
> >
> > What would an ideal approach look like?
> 
> My suggestion would be one of two ways:
>    1. use xsd:gYear in spite of the spec's restriction; or
>    2. use xsd:integer. Although gYear is formally disjoint from integer,
> in practice they are almost treated exactly in the same way.

OK. We will continue to use xsd:gYear, xsd:gYearMonth and xsd:date and hope that it works for our customers.

Best,

Lars

> >
> > [1] http://www.w3.org/TR/owl2-syntax/#Datatype_Definitions

Received on Friday, 27 June 2014 13:55:41 UTC