W3C home > Mailing lists > Public > public-lod@w3.org > January 2015

Re: Quick Poll

From: Stian Soiland-Reyes <soiland-reyes@cs.manchester.ac.uk>
Date: Mon, 26 Jan 2015 07:47:32 +0000
Message-ID: <CAPRnXtm5wUcbY99SQ=6VWEMBj8oEoGzRuNYZng2L40qJBYenXw@mail.gmail.com>
To: Hugh Glaser <hugh@glasers.org>
Cc: W3C LOD Mailing List <public-lod@w3.org>
I would have to check the VOIDs we consume, but I believe generally if you
are example.com, you try to publish linksets to other URIs with your
example.com/* as subjects - after all those are what you claim to be
authorative for.

The choice of an asymmetric property may mean sacrificing this principle if
there is no inverse, but often there is - for instance
prov:specializationOf vs prov:generalizationOf.

In Open PHACTS we used to calculate the inverse direction of symmetrical
properties at data loading time, but now we do that (and transitive
identifier mapping) on the fly at query time.

It turns out that when you ask domain scientists, they might disagree with
the owl:sameAs statements in one of the directions (!) - so this is now
configurable per dataset.
On 25 Jan 2015 12:28, "Hugh Glaser" <hugh@glasers.org> wrote:

> Thanks Stian and Alasdair,
> Just going back to the original question for a moment, I’ll try another
> way of putting it for people who don’t have their own URIs.
>
> When you want to create a set of links (of the sorts of properties you are
> talking about, but only the symmetric ones), you have often started with
> candidate URIs, and then found other URIs that have that relationship.
> When you create the triples to record this valuable information, does the
> original candidate appear as
> a) subject, b) object, or c) just whatever, or d) maybe you assert 2
> triples both ways?
>
> It’s a very qualitative and woolly question, I realise :-)
> Thanks.
>
>
> > On 25 Jan 2015, at 10:45, Stian Soiland-Reyes <
> soiland-reyes@cs.manchester.ac.uk> wrote:
> >
> > For properties you would need to use owl:equivalentProperty or
> rdfs:subPropertyOf in either direction.
> >
> > SKOS is very useful here as an alternative when the logic gets dirty due
> to loose term definition.
> >
> > As an example see this SKOS mapping from PAV onto Dublin Core Terms
> (which are notoriously underspecified and vague):
> >
> > http://www.jbiomedsem.com/content/4/1/37/table/T5
> >
> > http://www.jbiomedsem.com/content/4/1/37 (Results)
> >
> > Actual SKOS: http://purl.org/pav/mapping/dcterms
> >
> > Here we found SKOS as a nice way to do the mapping independently (and
> justified) as the inferences from OWL make DC Term incompatible with any
> causal provenance ontology like PROV and PAV.
> >
> > On 23 Jan 2015 17:59, "Hugh Glaser" <hugh@glasers.org> wrote:
> > Thanks, and thanks for all the answers so far.
> >
> > > On 23 Jan 2015, at 16:23, Stian Soiland-Reyes <
> soiland-reyes@cs.manchester.ac.uk> wrote:
> > >
> > > Not sure where you are going, but you are probably interested in
> > > linksets - as a way to package equivalence relations - typically in a
> > > graph of its own.
> > Thanks - I have a lot of linksets :-)
> > >
> > > http://www.w3.org/TR/void/#describing-linksets
> > >
> > >
> > >
> > > To answer the questions:
> > >
> > > Q1: d) in subject, property, object, or multiple of those.
> > I don’t understand where property comes in for using owl:sameAs (or
> whatever) in stating equivalence between URIs, so I’ll read that as c)
> > >
> > >
> > > Q2: No. We already reuse existing vocabularies and external
> > > identifiers, and there could be a nested structure which is only
> > > indirectly connected to "our" URIs.
> > I realise that this second question wasn’t as clear as it might have
> been.
> > What I meant was concerned with the sameAs triples only (as was explicit
> for Q1).
> > So, to elaborate, if you have decided that:
> > http://mysite.com/foo, http://dbpedia.org/resource/foo,
> http://rdf.freebase.com/ns/m.05195d8
> > are aligned (the same), then what do the triples describing that look
> like?
> > In particular, do you have any that look like
> > <http://dbpedia.org/resource/foo> owl:sameAs <
> http://rdf.freebase.com/ns/m.05195d8> .
> > (or vice versa), or do you equivalent everything to a “mysite” URI?
> >
> > But I guess for OpenPHACTS this doesn’t apply, since I understand from
> what you say below that you never mint a URI of your own where you know
> there is an external one.
> > Although it does beg the question, perhaps, of what you do when you
> alter find equivalences.
> >
> > Best
> > Hugh
> > >
> > > <http://example.com/our/own> pav:authoredBy
> > > <http://orcid.org/0000-0001-9842-9718> .
> > > <http://orcid.org/0000-0001-9842-9718> foaf:name "Stian
> Soiland-Reyes" .
> > >
> > > It's true you would also get the second triple from ORCID (remember
> > > content negotiation!), but it's very useful for presentation and query
> > > purposes to include these directly, e.g. in a VOID file.
> > >
> > > In most cases we do however not have any "our URIs" except for
> > > provenance statements. But perhaps Open PHACTS is special in that
> > > regard as we are integrating other people's datasets and shouldn't be
> > > making up any data of our own. :)
> > >
> > >
> > >
> > > Perhaps also of interest:
> > >
> > > In the Open PHACTS project <http://www.openphacts.org/> we use this
> > > extensively - we let the end-user choose which linksets of weak and
> > > strong equivalences they want to apply when a query is made. Such a
> > > collection of linksets and their application we call a "lense" - so
> > > you apply lenses to merge/unmerge your data. See
> > > http://www.slideshare.net/alasdair_gray/gray-compcoref
> > >
> > >
> > > In our identity mapping service
> > > <
> http://www.openphacts.org/about-open-phacts/how-does-open-phacts-work/identities-within-open-phacts
> >
> > > we pass in several parameters - the minimal is the URI to map.
> > >
> > > See http://openphacts.cs.man.ac.uk:9092/QueryExpander/mapURI and use
> > > http://rdf.ebi.ac.uk/resource/chembl/targetcomponent/CHEMBL_TC_2443 as
> > > the URI.
> > >
> > >
> > > We also have a piece of magic that can rewrite a SPARQL query to use
> > > the mapped URIs for a given variable (adding FILTER statements) try -
> > > http://openphacts.cs.man.ac.uk:9092/QueryExpander/
> > >
> > >
> > >
> > >
> > > On 23 January 2015 at 11:39, Hugh Glaser <hugh@glasers.org> wrote:
> > >> I would be really interested to know, please.
> > >> I suggest answers by email, and I’ll report back eventually.
> > >>
> > >> Here goes:
> > >> Imagine you have some of your own RDF using URIs on your base/domain.
> > >> And you have reconciled some of your URIs against some other stuff,
> such as dbpedia, freebase, geonames...
> > >> Now, visualise the owl:sameAs (or skos:whatever) triples you have
> made to represent that.
> > >>
> > >> Q1: Where are your URIs?
> > >> a) subject, b) object, c) both
> > >> Q2: Do all the triples have one of your URIs in them?
> > >> a) yes, b) no
> > >>
> > >> It’s just for a choice I have about the input format for sameAs
> services, so I thought I would ask :-)
> > >>
> > >> Best
> > >> Hugh
> > >> --
> > >> Hugh Glaser
> > >>   20 Portchester Rise
> > >>   Eastleigh
> > >>   SO50 4QS
> > >> Mobile: +44 75 9533 4155, Home: +44 23 8061 5652
> > >>
> > >>
> > >>
> > >
> > >
> > >
> > > --
> > > Stian Soiland-Reyes, eScience Lab
> > > School of Computer Science
> > > The University of Manchester
> > > http://soiland-reyes.com/stian/work/
> http://orcid.org/0000-0001-9842-9718
> > >
> >
> > --
> > Hugh Glaser
> >    20 Portchester Rise
> >    Eastleigh
> >    SO50 4QS
> > Mobile: +44 75 9533 4155, Home: +44 23 8061 5652
> >
> >
>
> --
> Hugh Glaser
>    20 Portchester Rise
>    Eastleigh
>    SO50 4QS
> Mobile: +44 75 9533 4155, Home: +44 23 8061 5652
>
>
>
Received on Monday, 26 January 2015 07:48:01 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 16:22:18 UTC