[Patterns] Annotation (was Re: Announce: Linked Data Patterns book)

Hi,

Have changed the subject line to clarify the discussion.

On 7 April 2010 00:14:15 UTC+1, Peter Ansell <ansell.peter@gmail.com> wrote:
> In the Annotation publishing pattern section there is the following statement:
>
> "It is entirely consistent with the Linked Data principles to make
> statements about third-party resources."
>
> I don't believe that to be true, simply because, unless users are
> always using a quad model (RDF+NamedGraphs), they have no way of
> retrieving that information just by resolving the foreign identifier
> which is the subject of the RDF triple.

Yes, the data won't be directly accessible from the original URI. But
it may be indirectly accessible by following an explicit link to the
Annotation data (the See Also pattern) introduced by the publisher, or
through the use of additional web services, e.g. Sindice, etc.

My assertion is that so long as the data can be found by following
hypertext links, then the data is still part of the overall web of
data. Clearly the downside is that there are some overheads (publisher
adding links; use of extra sources). The pattern description should
make this more explicit. btw, I wrote a blog post about this topic
recently [1].

I think there are trade-offs with both approaches:

* Mint new URIs -- requires more infrastructure (to publish), requires
reasoning (to merge), but data is discoverable by de-referencing
* Annotate -- easier to publish, merges easily, but requires more
infrastructure or co-ordination for data to be discoverable

> If the idea of the book is to point to quads as the preferred model
> for publishing Linked Data then it may be okay because the named graph
> is in each statement to indicate where the statement came from so
> users just have to recognise that URIs in triples are not linked to
> retrieval locations, and they should keep the quad extension and use
> that for retrieval. It seems like a stretch to have this as a valid
> alternative instead of the easier (in Linked Data terms) pattern of
> creating new URI's that are equivalent to the other URI, just with one
> or more extra statements.

I don't really see how Named Graphs are relevant here. And I'm
certainly not advocating that everyone has to publish their data in
that way.

> In the Equivalent Links publishing pattern section there is the
> following statement:
>
> "The relations allow for more fuzzy notions of equivalence and have
> weaker semantics: skos:exactMatch declares two concepts to be the same
> but doesn't imply that all statements about one concept are also true
> of another."
>
> Should indicate there that SKOS implies that all of the URIs are
> philosophically just concepts such as those that exist in taxonomies.
> To use skos:exactMatch one has to follow the SKOS model through or it
> doesn't fit, as with the OWL model.

Yes, good point. Understanding proper role of SKOS and OWL is
something that needs addressing.

Cheers,

L.

[1]. http://www.ldodds.com/blog/2009/12/annotated-data/


-- 
Leigh Dodds
Programme Manager, Talis Platform
Talis
leigh.dodds@talis.com
http://www.talis.com

Received on Wednesday, 7 April 2010 08:55:37 UTC