search

ec-geolink / design

Design information about the EarthCube Geolink project.
8 stars 1 forks source link

Vocabularies: Discuss practice of minting GeoLink URIs for SKOS concepts #74

Open amoeba opened 8 years ago

amoeba commented 8 years ago

Looking at http://schema.geolink.org/voc/index.html I see a number of vocabularies like those from NERC that are being mapped from SKOS concepts into GeoLink as OWL classes. I see in at least BCO-DMO's graphs, these minted GeoLink URIs are being used like:

  <rdf:Description rdf:about="http://data.geolink.org/id/bcodmo/instrument/404">
    <owl:sameAs rdf:resource="http://schema.geolink.org/dev/voc/nvs/L05#LAB20"/>
  ...

where http://schema.geolink.org/dev/voc/nvs/L05#LAB20 is 

    <owl:Class rdf:about="http://schema.geolink.org/dev/voc/nvs/L05#LAB20">
        <rdfs:subClassOf rdf:resource="http://schema.geolink.org/dev/base/main#Instrument"/>
    </owl:Class>

    <rdf:Description rdf:about="http://schema.geolink.org/dev/voc/nvs/L05#LAB20">
        <rdfs:label xml:lang="en">spectrophotometers</rdfs:label>
        <rdfs:seeAlso rdf:resource="http://vocab.nerc.ac.uk/collection/L05/current/LAB20/"/>
    </rdf:Description>

What is the goal of importing these SKOS concepts from the NERC vocabularies in this way? Can we avoid this duplication and reference the NERC vocabularies directly?

mbjones commented 8 years ago

@mpsaloha and @dr-shorthair raised these questions to us recently and we didn't have good answers. Seems like we could use the NERC URIs directly, rather than through the very weak rdfs:seeAlso relation. Currently I can't see any advantages to the indirection, but there is the disadvantage of weakening the semantic relationship to the NERC vocabularies.

krisnadhi commented 8 years ago

If I remember correctly, that choice was because in GeoLink we did assert the relationship between vocabulary terms into subclass/subsumption for those that refer to platform types or instrument types. Note that these subsumptions are fairly obvious, at least to me, as a modeler, although their counterparts in NERC only use SKOS. I believe @robertarko was reluctant to assert subsumptions directly on the NERC vocabulary because we are not its maintainer.

dr-shorthair commented 8 years ago

Ah - the NIH syndrome ;-) ("not invented here" not national institutes for health)

I know the team that runs the NERC Vocabulary Service (so does @robertarko) and my sense is that the NVS is likely to be highly reliable and persistent. They haven't been the most modern in terms of their embrace of semantic technologies, but they have a 15 year legacy of assembling and curating vocabularies to protect. It had been in Oracle, and Roy Lowry is a PL/SQL whizz. So the SKOS based system was just an export from Oracle.

But even Roy has seen the light in recent months and I think had comitted to an RDF environment moving forward. Furthermore, Roy has now officially retired, and the new guard (Alexandra Kokkinaki, Rob Thomas) have been tasked to move it in that direction.

The primary concern, however, is probably (i) are the URIs stable (ii) are the subsumption relationships between concepts stable (regardless of whether they are expressed using skos:broader or rdfs:subClassOf )

Then there are all the other relationships that are being progressively built in to the NVS system, as they develop richer ontologies (rather than just SKOS).

My rather strong recommendation is to accept NVS as what it is - one of the most stable, best curated resources out there, furthermore which is hosted by an operational organization (the BODC within NERC) and not just in the context of a single project that is planned to go way in a small number of years. NVS is already widely used in the marine community (where it originated) and is mission critical for a sequence of European initiatives, so is unlikely to be allowed to go away. I'd rather see us providing feedback back to NVS encouraging them to further refine their practices to allow it to be used as a semantically reliable resource, than set up temporary islands elsewhere.

dr-shorthair commented 8 years ago

If we really want to move up from isolated projects to the semantic web, we have to trust some systems out on the web that we don't actually control, and have a plan for how to work around issues that come up, rather than retreat into a walled garden.

dr-shorthair commented 8 years ago

FWIW - here's a summary of what is in NVS - http://goo.gl/zGMVuf

bob-arko commented 8 years ago

We had a breakthrough discussion on exactly this topic, at the ODIP Workshop in Paris a few weeks ago. Adam Leadbetter showed an example of publishing the NVS vocabulary terms as both SKOS concepts and OWL individuals, using an overlay approach that would preserve the existing URIs. (I may not have the exact details correct. But you get the idea.)

I'd love to use the NVS URIs natively, instead of republishing at geolink.org. Big win, both practical and political.

On Thu, Oct 15, 2015 at 05:44:28PM -0700, Simon Cox wrote:

If we really want to move up from isolated projects to the semantic web, we have to trust some systems out on the web that we don't actually control, and have a plan for how to work around issues that come up, rather than retreat into a walled garden.

Reply to this email directly or view it on GitHub: https://github.com/ec-geolink/design/issues/74#issuecomment-148562175

dr-shorthair commented 8 years ago

publishing... as both SKOS concepts and OWL individuals

For http://environment.data.gov.au/def/property/ we asserted qudt:QuantityKind rdfs:subClassOf skos:Concept .

and then things like http://environment.data.gov.au/def/property/arsenic_concentration are simultaneously a skos:Concept as well as an individual from the class qudt:QuantityKind, which has much richer semantics. Is that what you mean?

krisnadhi commented 8 years ago

@robertarko do you still have Adam Leadbetter's example that you mentioned?

bob-arko commented 8 years ago

See below, example of the NVS P02 term PSAL (Salinity of the water column).

Would allow incorporation of OWL without requiring new URIs. Said he ran it through HermiT and FaCT++.

@prefix skos: http://www.w3.org/2004/02/skos/core#. @prefix owl: http://www.w3.org/2002/07/owl#.

http://vocab.nerc.ac.uk/collection/P02/current/ a skos:Collection; skos:member http://vocab.nerc.ac.uk/collection/P02/current/PSAL/.

http://vocab.nerc.ac.uk/collection/P02/current/PSAL/ a skos:Concept; skos:prefLabel "Salinity"@en; skos:narrower http://vocab.nerc.ac.uk/collection/P01/current/PSALCU01/; a owl:Class.

http://vocab.nerc.ac.uk/collection/P01/current/ a skos:Collection; skos:member http://vocab.nerc.ac.uk/collection/P01/current/PSALCU01/.

http://vocab.nerc.ac.uk/collection/P01/current/PSALCU01/ a skos:Concept; skos:prefLabel "Practical salinity in the water column"@en; a http://vocab.nerc.ac.uk/collection/P02/current/PSAL/; skos:broader http://vocab.nerc.ac.uk/collection/P02/current/PSAL/.

On Wed, Nov 18, 2015 at 10:45:21AM -0800, krisnadhi wrote:

@robertarko do you still have Adam Leadbetter's example that you mentioned?