W3C home > Mailing lists > Public > public-openannotation@w3.org > February 2013

Re: Semantic Tags (was several threads)

From: Paolo Ciccarese <paolo.ciccarese@gmail.com>
Date: Mon, 4 Feb 2013 12:49:32 -0500
Message-ID: <CAFPX2kD0r6C+O0XAeVTWQLnrNv5E8HLea8GDLNtFTJ6PSZdjWA@mail.gmail.com>
To: Robert Sanderson <azaroth42@gmail.com>
Cc: Antoine Isaac <aisaac@few.vu.nl>, public-openannotation@w3.org
On Mon, Feb 4, 2013 at 12:18 PM, Robert Sanderson <azaroth42@gmail.com>wrote:

> To try to pull the threads together ...
>
> Issue:  If there is a document which an annotator wants to use as a
> semantic tag, then it is not possible to say that it's an oa:Tag, as
> that information is specific to the Annotation.
>
> Use cases: Many use cases, especially in bioinformatics.
>
> Severity:  Difficult to determine and somewhat mitigated by the
> (unanimous?) consensus that it is bad modeling and against the
> architecture of the WWW to have a URI identify both a concept and a
> document at the same time.  Severe enough in communities that need it
> that it would be great if it was addressed.
>
> Current:  The spec does not say exactly how to solve the problem, but
> recommends minting a new URI for the tag and relating it "somehow" to
> the document. It also has a single oa:Tag class, and relies on the
> presence or non-presence of cnt:chars.
>
> Regarding, first oa:Tag versus oa:SemanticTag:
>
> * The open world assumption means that the non-presence of cnt:chars
> means "we don't know if it's a semantic tag or not".
> * It's not our predicate to associate additional semantics with its
> presence, or lack thereof
> * If you get an HTTP URI that calls itself a tag, and has cnt:chars,
> it's unclear what to do.
>
> Thus the proposal is to have a subclass, oa:SemanticTag to avoid these
> situations.
>
>
> There are several implicit proposals as to the model, all of which
> further clarify the current recommendation:
>
> 1.  (Rob) Use Specific Resource with a oa:SemanticTag class. Then the
> object of oa:hasSource is the document.  Objection from Antoine: This
> is abusing Specific Resources.
>
> 2.  (Antoine) Use a oa:SemanticTag class, with foaf:primaryTopicOf.
> Object from Rob: it's inverse functional, so the same document
> couldn't be used for different semantic concepts. As the URI for the
> tag resource is likely going to be a UUID or a blank node, this could
> have unfortunate repercussions.
>

-1 The 'inverse functional' constraint is too restrictive


>
> 3.  (Rob) Use oa:SemanticTag class, with foaf:page.  This is the same
> as 2. but with a looser predicate that isn't functional.
>

The last one is compact, does not interfere with other constructs, gives a
little structure without too much commitment, is more declarative.

And I like how it reads for cases in which the URI is actually a page or
HTML document
ex:semtag a oa:SemanticTag ;
  foaf:page <http://omim.org/entry/104760> .

The inverse I think also makes sense:
<http://omim.org/entry/104760> foaf:topic ex:semtag

However, for URis such as the DBpedia ones, are we still planning to do:
<http://dbpedia.org/resource/Eiffel_Tower> a oa:SemanticTag. ?
or to adopt the above model?

Paolo
Received on Monday, 4 February 2013 17:49:59 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 16:38:22 UTC