Correction annotations with closely coupled bodies and targets?

Several frequently mentioned annotation use cases make use of annotations
that identify a specific correction to or enrichment of a resource. These
annotations are particularly useful when doing shared curation of digital
collections - e.g., correcting OCR'd text held in a collection of digitized
books, marking transcribed text in a digitized novel as speech, etc. In
these contexts, annotations are often designed to be actionable - e.g.,,
XSLT prescribing how to correct the TEI resource annotated.

 

Consider annotating a small error in a TEI transcript of a digitized book.
How might the XSLT to make this correction best be described as an
annotation body? (Or is this even the right way to describe this kind of
annotation?)  For such an annotation, how might the description of an XSLT
as annotation body compare to description of the target to which it is meant
to be applied? Is the following example a viable approach:

 

<http://example.org/myAnnotation1> a oa:Annotation  ;

        oa:hasBody <urn:uuid:1111...>  ;

        oa:hasTarget <urn:uuid:2222...> .

 

<urn:uuid:1111...> a oa:SpecificResource;

        oa:hasSource <http://myUniversity.edu/Page1Transcript.xml> ;

        oa:hasSelector < urn:uuid:3333...> .

 

<urn:uuid:3333...> a oa:FragmentSelector ;

        rdf:value "...[XPointer fragment identifier]..." .

<urn:uuid:2222...> a oa:SpecificResource ;

        oa:hasSource <http://myUniversity.edu/Page1Transcript.xml> ;

        oa:hasSelector < urn:uuid:3333...> ; 

        oa:hasStyle <http:/myUniversity.edu/myXSLT.xsl>.

 

<http:/myUniversity.edu/myXSLT.xsl> a oax: oax:XsltStyle ;

        rdf:type " application/xslt+xml" .   (or perhaps "text/xsl")

 

Presumably could also embed the XSLT directly in the annotation using
content in RDF as shown in the extension namespace for CSS applied to a
specifiedTarget.

 

Note that both the body (<urn:uuid:1111...>) and target <urn:uuid:2222...>
of this simple annotation are of class oa:SpecificResource. Both the body
and the target share the same oa:hasSource triple and the same
oa:FragmentSelector. 

 

The difference is the additional oa:hasStyle for the body. (Section 5.3 of
the current Open Annotation core model suggests that XSLT is a specific
format that can be used as object of a oa:hasStyle predicate.) 

 

If you apply the FragmentSelector to the Source, you get the TEI fragment
you want to correct.

 

If you apply the FragmentSelector to the Source and then apply the Style
(the XSLT), you get the TEI fragment corrected as you want it to be. 

 

These seem logical as target and body (respectively) for an annotation meant
to point out an error and express how it should be corrected (i.e., what
should replace it). 

 

Is this usage consistent with the intent of the Open Annotation model? If
so, is it good practice? Are there other (better) ways to describe this
annotation? How does the logic of such an annotation relate to the logic of
a bodiless annotation that has an oa:hasStyle attribute for expressing the
highlighting or text color on the specifiedTarget?

 

Feedback appreciated.

 

Tim Cole

University of Illinois at Urbana-Champaign