W3C home > Mailing lists > Public > public-vocabs@w3.org > January 2013

Re: Should we adopt SKOS?

From: Martin Hepp <martin.hepp@ebusiness-unibw.org>
Date: Thu, 10 Jan 2013 12:23:52 +0100
Cc: Dan Brickley <danbri@danbri.org>, Richard Wallis <richard.wallis@oclc.org>, "public-vocabs@w3.org" <public-vocabs@w3.org>, Jamie Taylor <jamietaylor@google.com>
Message-Id: <4CC5D78E-FAB9-4E15-8547-7835651772D2@ebusiness-unibw.org>
To: Ivan Herman <ivan@w3.org>
Hi Ivan:

For the non-trivial cases of mappings, I personally prefer SPARQL CONSTRUCT rules over anything else. We will also add those to the documentation of GoodRelations for the few cases in which schema.org supports now two patterns, shallow, original schema.org elements and more generic GoodRelations elements (e.g. opening hours or pricing).

A caveat is that except for SPIN, there is AFAIK no standard syntax for exchanging such SPARQL CONSTRUCT rules and tying them to the relevant vocabulary.

We currently just publish them at

    http://wiki.goodrelations-vocabulary.org/Axioms

Martin

On Jan 10, 2013, at 12:08 PM, Ivan Herman wrote:

> Hi Martin,
> 
> comments below
> 
> On Jan 10, 2013, at 11:15 , Martin Hepp <martin.hepp@ebusiness-unibw.org> wrote:
> 
>> Hi Ivan, all:
>> 
>> FYI: For GoodRelations, we will provide a formal, OWL/RDFS-based mapping in the next GR service update, i.e. for an RDF client who can handle owl:sameAs, rdfs:subClassOf, rdfs:subPropertyOf, owl:equivalentProperty, and owl:equivalentClass (in a moderate, pragmatic way - no need for full OWL reasoning),
> 
> Agreed. My take is that OWL RL should be the maximum level of OWL we aim for (although, see some comments below).
> 
>> it will not matter whether GoodRelations is used in the original namespace or in the schema.org namespace.
>> 
>> E.g.
>> 
>> @prefix schema: <http://schema.org/> .
>> @prefix gr: <http://purl.org/goodrelations/v1#>.
>> 
>> schema:ParcelService a owl:Class ;
>> 	owl:equivalentClass gr:DeliveryModeParcelService .
>> 
>> By that, we can also easily bridge the slight differences in naming conventions between the two namespaces, as documented in [1]
>> 
> 
> Yes, this simple structure may cover GR. However, isn't it possible/feasible that, in some cases, the mapping is not that straightforward? There may be, eg, context-sensitive subproperty mappings; ie, if a property is used on the instance of a class _then_, but only then is there a, say, subproperty relationship. I hope that n3 rule below is understandable:
> 
> { ?x a ex:A. ?x ex:p ?y } => { ?x ex:q ?y}
> 
> The reason this may arise is because schema, for obvious reasons, tries to minimize the number of terms it uses, so two different vocabularies may end up using the same property. The relationship to the 'original' one may depend on the 'context', eg, the class/type of the resource.
> 
> I actually had a discussion with an OWL expert (Ian Horrocks) and he showed me a way to express this rule using OWL2 RL (I was actually surprised it was possible) but it is convoluted and not for the faint hearted:-). Ie, I am not sure we want to go along that way (maybe yes) and, also, whether all cases coming up would be transferable to OWL RL.
> 
> Maybe, pragmatically, we would ignore these. Maybe we would use some rule formalism (n3, sparql, whatever). But we have to know about this.
> 
>> For the near future, I think that defining such a thing as a "best practice" should be sufficient. We could then put the actual task on the shoulders of the maintainers of the original vocabulary.
> 
> + having a pointer at those files at schema.org. So that they could be easily found... (eg, in the OWL schema.org vocabulary as a seealso, or something similar).
> 
>> 
>> However, should schema.org evolve beyond ca. 1000 types and include components from more than ca. 10 vocabularies that are not fully stable, we will need a more flexible and scalable approach.
>> 
>> Ideally, we would already sketch the architecture for this future now (nice PhD or research paper topic) while we muddle through with the pragmatic approach sketched above.
> 
> Yep
> 
> Thanks
> 
> Ivan
> 
>> 
>> Martin
>> 
>> [1] http://wiki.goodrelations-vocabulary.org/Cookbook/Schema.org#Naming_Differences
>> 
>> 
>> On Jan 10, 2013, at 11:00 AM, Ivan Herman wrote:
>> 
>>> Exactly... 
>>> 
>>> I think it would worthwhile to have some general discussion on how to maintain equivalence between what you call 'adopted' vocabularies and their original. We do not want to disenfranchise users who happen to use (or have used) the original namespace/terms.
>>> 
>>> Putting my RDF hat on, it would/could be possible to set up some machine readable files defining the equivalences although, I am afraid, a simple owl:sameAs, owl:samePropertyAs, etc, may not work in all cases. There is also an issue of who would manage those equivalences: is it the job of schema.org? The 'authors/owners' of the original vocabularies? This group as a community? I honestly do not know. And I am also not 100% whether this is the right approach.
>>> 
>>> At the moment we have two 'adopted' vocabularies (GoodRelations and rNews); out of those GR has a significant adoption already in its original namespace, I am not sure about the rNews figures. LRMI, afaik, does not have a non-schema version, so that is not an issue. But if we begin to discuss SKOS, SIOC and the others, then it is worthwhile having this discussion before it gets out of hand...
>>> 
>>> Whether it is a workshop or simply a discussion on this list: I do not know. I would certainly like to hear the opinions on this list.
>>> 
>>> Thanks Martin, for raising this.
>>> 
>>> Ivan
>>> 
>>> On Jan 9, 2013, at 19:28 , Martin Hepp <martin.hepp@ebusiness-unibw.org> wrote:
>>> 
>>>> I would like to add at this point that I see schema.org first and foremost as an effort to provide ex ante schema alignment under a common, easily accessible umbrella, following a "schema alignment before data publication" paradigm.
>>>> This makes authoring markup easier (no namespace traversals) and reduces the complexity for consuming clients.
>>>> 
>>>> This having said, I would also rather suggest to import a pragmatic subset of SKOS into schema.org, in a fashion similar to the external enumerations pattern in use now and the additionalType property solution for external type systems.
>>>> 
>>>> An general open question is how we maintain the alignment of the original vocabularies (LRMI, GoodRelations, maybe SKOS, SIOC, ...) with their conceptually equivalent elements in schema.org.
>>>> 
>>>> In GoodRelations, we will simply try to keep both in sync manual, which seems reasonable given that the conceptual model is pretty stable.
>>>> 
>>>> In the long run, however, we will need to find a solution that combines the centralized schema.org approach with distributed contributions from various domains.
>>>> 
>>>> Maybe that is an interesting topic for a new schema.org workshop in 2013?
>>>> 
>>>> Best
>>>> 
>>>> Martin
>>>> 
>>>> 
>>>> 
>>>> On Jan 9, 2013, at 7:00 PM, Dan Brickley wrote:
>>>> 
>>>>> +Cc: Jamie
>>>>> 
>>>>> On 9 January 2013 16:29, Richard Wallis <richard.wallis@oclc.org> wrote:
>>>>>> Coming from the bibliographic world, specifically chairing  the Schema Bib
>>>>>> Extend Group[1] (who are building a consensus around a group of proposals
>>>>>> for Schema.org extensions for bibliographic resources, before submitting
>>>>>> them to this group), I am identifying situations where being able to model
>>>>>> things as SKOS[2] Concepts held in ConceptSchemes would make a great deal of
>>>>>> sense.
>>>>>> 
>>>>>> Working with colleagues we were finding ourselves almost reinventing the
>>>>>> SKOS model in [proposed] Schema.org vocabulary.
>>>>>> 
>>>>>> The introduction of External Enumerations[2] provided the ability to link to
>>>>>> lists of things controlled by external authorities.  An approach used widely
>>>>>> in the bibliographic and other domains  Library of Congress Subject
>>>>>> Headings[4] for example.  Many of these authorities are modelled using SKOS
>>>>>> (Concepts within ConceptSchemes) which introduces a consistent structured
>>>>>> way to describe relationships (broader/narrower), language specific
>>>>>> preferred labels, etc.
>>>>>> 
>>>>>> Sub-typing Intangible for Concept and ConceptScheme, it would be
>>>>>> comparatively easy to introduce SKOS into Schema.  The benefits I believe
>>>>>> being to add even more value to External Enumeration; providing a flexible
>>>>>> simple-ish yet standard pattern for marking up lists of concepts and their
>>>>>> interrelationships; provide a very easy way for already published
>>>>>> authoritative lists of concepts to adopt Schema.org and provide valuable
>>>>>> resources for all to connect with.
>>>>>> 
>>>>>> For instance VIAF[4] the Virtual International Authority File, a well used
>>>>>> source of URIs and authoritative names for people and organisations
>>>>>> (compiled and managed by the bibliographic community but used widely) is
>>>>>> already in SKOS.  SKOS is also used in many other domains.
>>>>>> 
>>>>>> I could see this adding value without significant impact on the rest of
>>>>>> Schema.
>>>>>> 
>>>>>> What do others think?
>>>>> 
>>>>> I've been thinking in this direction too (and had brief discussion
>>>>> with Jamie, cc:'d, w.r.t. Freebase's approach).
>>>>> 
>>>>> SKOS has done well and a great many controlled vocabularies in the
>>>>> thesauri, subject classification and code list tradition are expressed
>>>>> using it. SKOS handles various cases where 'class/object/property'
>>>>> models don't capture things well. I'd like to have a way of reflecting
>>>>> SKOS-oriented data into schema.org descriptions without going
>>>>> 'multi-namespace'. There are also already various corners of
>>>>> schema.org where different loose notions of 'category' are slipping
>>>>> in.
>>>>> 
>>>>> My current preference would be to call a new type "Topic" or perhaps
>>>>> "Category" rather than the more esoteric / vague "Concept", even while
>>>>> borrowing most structure and terminology from SKOS.
>>>>> 
>>>>> Do you have a strawman list of what you'd hope to include, from a
>>>>> bibliographic perspective?
>>>>> 
>>>>> Dan
>>>>> 
>>>>>> ~Richard
>>>>>> 
>>>>>> --
>>>>>> Richard Wallis
>>>>>> Technology Evangelist
>>>>>> OCLC
>>>>>> 
>>>>>> 
>>>>>> 
>>>>>> [1] http://www.w3.org/community/schemabibex/
>>>>>> [2] http://www.w3.org/TR/skos-reference/
>>>>>> [3] http://www.w3.org/wiki/WebSchemas/ExternalEnumerations
>>>>>> [4] http://id.loc.gov/authorities/subjects.html
>>>>>> 
>>>>> 
>>>> 
>>>> --------------------------------------------------------
>>>> martin hepp
>>>> e-business & web science research group
>>>> universitaet der bundeswehr muenchen
>>>> 
>>>> e-mail:  hepp@ebusiness-unibw.org
>>>> phone:   +49-(0)89-6004-4217
>>>> fax:     +49-(0)89-6004-4620
>>>> www:     http://www.unibw.de/ebusiness/ (group)
>>>>      http://www.heppnetz.de/ (personal)
>>>> skype:   mfhepp 
>>>> twitter: mfhepp
>>>> 
>>>> Check out GoodRelations for E-Commerce on the Web of Linked Data!
>>>> =================================================================
>>>> * Project Main Page: http://purl.org/goodrelations/
>>>> 
>>>> 
>>>> 
>>>> 
>>> 
>>> 
>>> ----
>>> Ivan Herman, W3C Semantic Web Activity Lead
>>> Home: http://www.w3.org/People/Ivan/
>>> mobile: +31-641044153
>>> FOAF: http://www.ivan-herman.net/foaf.rdf
>>> 
>>> 
>>> 
>>> 
>>> 
>> 
>> --------------------------------------------------------
>> martin hepp
>> e-business & web science research group
>> universitaet der bundeswehr muenchen
>> 
>> e-mail:  hepp@ebusiness-unibw.org
>> phone:   +49-(0)89-6004-4217
>> fax:     +49-(0)89-6004-4620
>> www:     http://www.unibw.de/ebusiness/ (group)
>>        http://www.heppnetz.de/ (personal)
>> skype:   mfhepp 
>> twitter: mfhepp
>> 
>> Check out GoodRelations for E-Commerce on the Web of Linked Data!
>> =================================================================
>> * Project Main Page: http://purl.org/goodrelations/
>> 
>> 
>> 
>> 
> 
> 
> ----
> Ivan Herman, W3C Semantic Web Activity Lead
> Home: http://www.w3.org/People/Ivan/
> mobile: +31-641044153
> FOAF: http://www.ivan-herman.net/foaf.rdf
> 
> 
> 
> 
> 

--------------------------------------------------------
martin hepp
e-business & web science research group
universitaet der bundeswehr muenchen

e-mail:  hepp@ebusiness-unibw.org
phone:   +49-(0)89-6004-4217
fax:     +49-(0)89-6004-4620
www:     http://www.unibw.de/ebusiness/ (group)
         http://www.heppnetz.de/ (personal)
skype:   mfhepp 
twitter: mfhepp

Check out GoodRelations for E-Commerce on the Web of Linked Data!
=================================================================
* Project Main Page: http://purl.org/goodrelations/
Received on Thursday, 10 January 2013 11:24:17 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 10 January 2013 11:24:17 GMT