Re: Annotation entailment!

Sorry - I am confused now. Time for a concrete use-case. I have found myself 
converting human-readable specs into OWL more than once. It is natural to 
want to annotate parts of the OWL as comming from parts of the spec 
documents. The granularity for this is at the owl1.1 axiom level, annotating 
these with the location in the source document that states the knowledge they 
capture. I'm usually quite careful to use axioms in the OWL that are as close 
as possible a direct translation of what is stated in the original spec, even 
if there are other potentially more OWL-friendly ways to say it. The 
semantics of the annotation I had assumed where that they applied to that 
axiom, not to the set of items with identical interpretations.

In the light of what is said below, should I not be doing this with 
annotations? To be clear, my intent was to capture a logical constraint, and 
*that exact* way of stating it, and associating this with the source 
reference.

Matthew (perplexed)

On Wednesday 04 July 2007, Jeremy Carroll wrote:
> Summary: why is this an issue? What are you trying to annotate, and why?
> Why are you using this mechanism?
>
> ======================
>
>
> The question is why are you using rdfs:label or eg:lastModifiedOn or
> whatever rather than <!--  XML Comments -->.
>
> If you want an annotation that is entirely invisible, except in an
> editor, the XML specification provides that mechanism:
> <!--
>
> -->
>
> If you use some other mechanism, then it is in order to get additional
> functionality.
>
> The functionality that OWL DL annotations provide is:
> - the annotation is part of the RDF graph
> - some basic semantics is provided.
> - the annotation is regarded as an annotation of the interpretation of
> the items in the graph, rather than an annotation of the graph syntax.
>
>
> i.e.
>
> There is a mechanism to annotate the XML syntax: XML comments.
>
> There is a mechanism to annotate the individuals properties and classes
> in the ontology.
>
> There is no mechanism to annotate the graph syntax.
>
> ===
>
> If a mechanism for annotating the graph syntax is desired, one method
> would be to create a new annotation property eg:annotatedGraph that
> takes an RDF/XML literal as its object.
>
> If your ontology is a set of triples O with name U, and O includes
> U rdf:type owl:Ontology
>
> then set
>
> O' = O union
> { U eg:annotatedGraph X^rdf:XMLLiteral }
>
> where X is a serialization of O, including appropriate XML comments.
>
> This provides a simple mechanism allowing annotations of the graph as a
> graph, (i.e. reading in, and writing it out does not loose the comments,
> but the comments have [no/vanishing little] semantic force).
>
> Jeremy

Received on Thursday, 5 July 2007 21:02:59 UTC