- From: Antoine Isaac <Antoine.Isaac@KB.nl>
- Date: Sat, 24 Jan 2009 14:44:33 +0100
- To: "Neubert Joachim" <J.Neubert@zbw.eu>, "Houghton,Andrew" <houghtoa@oclc.org>, <public-esw-thes@w3.org>
- Message-ID: <68C22185DB90CA41A5ACBD8E834C5ECD053E12B0@goofy.wpakb.kb.nl>
Hi Joachim, I'm not a specialist of these issues, but I'm actually quite sceptical about that principle in the case of "subject". Clearly, there is assymetry between the two directions of "subject" when one wants to "follow one's noise" (and I think this was mentioned when we discussed the skos:subject issue a while ago). Just imagine for the sake of example, that lcsh.info had been augmented with a linked data version of the catalogue of the Library of Congress: - when you're accessing a document, you are pointed to a small number of subjects this document is about. And quite naturally, you may want to follow your nose to these subjects, which is fine; - when you're accessing a concept, then quite often it will be the subject of dozens, if not thousands, of documents. Would it be natural then to provide them directly to the user? It might be rather confusing, and having this "direction" mediated by a query (and some more complex interface machinery that would go with it) seems to me more adequate. Antoine -------- Message d'origine-------- De: Neubert Joachim [mailto:J.Neubert@zbw.eu] Date: ven. 23/01/2009 15:39 À: Antoine Isaac; Houghton,Andrew; public-esw-thes@w3.org Objet : AW: RE : substitute for skos:isSubjectOf Hi Antoine, I agree with you that it's perfectly possible to query a SPARQL server either way round. In a closed world, there would be no problem. But the usage I want to encourage is the "follow-your-nose" principle Ed Summers described for the web of open linked data. This modus operandi makes heavy use of the fact that RDF forms a directed graph. Navigation through this graph in a straightforward way is possible only in the direction given by the existing triples. And with dc:subject (or any other relation that puts the concept in the object position), it is the wrong direction. So, when you are at the node of the concept, you come to nowhere without having the inverse relation materialized. - I hope this may explain my stubbornness about this issue ... Cheers, Joachim ________________________________ Von: Antoine Isaac [mailto:Antoine.Isaac@KB.nl] Gesendet: Freitag, 23. Januar 2009 11:32 An: Houghton,Andrew; Neubert Joachim; public-esw-thes@w3.org Betreff: RE : substitute for skos:isSubjectOf Dear Joachim, First a quick question: why would you be so upset about descriptor in the object position? In RDF --especially when you want to query an RDF graph with SPARQL-- there is no fundamental difference between having asserted a triple (r1,p,r2) and the triple (r2,p',r1), when p' is the inverse of p. (apart of course if one of your resources is a literal, but that's not your case) Cheers, Antoine -------- Message d'origine-------- De: public-esw-thes-request@w3.org de la part de Houghton,Andrew Date: jeu. 22/01/2009 18:27 À: Neubert Joachim; public-esw-thes@w3.org Objet : RE: substitute for skos:isSubjectOf Yes, that's what I meant. dct:relation says that there is a relationship between the skos:Concept and another URI given by rdf:Description. The dct:subject is a property of the URI given by rdf:Description. One could have split this out to be: <rdf:Description rdf:about="http://www.econis.eu/...=Black+market"> <dct:subject rdf:resource="http://zbw.eu/stw/descriptor/19289-4" /> </rdf:Description> <skos:Concept rdf:about="http://zbw.eu/stw/descriptor/19289-4"> <skos:prefLabel>Black market</skos:prefLabel> <dct:relation rdf:resource="http://www.econis.eu/...=Black+market" /> </skos:Concept> and if RDF encountered another definition some where in your data or any data you merged with, such as: <rdf:Description rdf:about="http://www.econis.eu/...=Black+market"> <dct:title>Black market search</dct:title> </rdf:Description> it would merge the dct:title and dct:subject under the http://www.econis.eu/...=Black+market URI. Andy. From: Neubert Joachim [mailto:J.Neubert@zbw.eu] Sent: Thursday, January 22, 2009 12:13 PM To: Houghton,Andrew; public-esw-thes@w3.org Subject: AW: substitute for skos:isSubjectOf So, the example you built would look like this: <skos:Concept rdf:about="http://zbw.eu/stw/descriptor/19289-4 <http://zbw.eu/stw/descriptor/19289-4> "> <skos:prefLabel>Black market</skos:prefLabel> <dct:relation> <rdf:Description rdf:about="http://www.econis.eu/...=Black+market <http://www.econis.eu/...=Black+market> "> <dct:subject rdf:resource="http://zbw.eu/stw/descriptor/19289-4" /> </rdf:Description> </dct:relation> </skos:Concept> Right, so far? The meaning, as I understand it, would be two statements about two different subjects: http://zbw.eu/stw/descriptor/19289-4 has a dct:relation to http://www.econis.eu/...=Black+market, and http://www.econis.eu/...=Black+market has a dct:subject of http://zbw.eu/stw/descriptor/19289-4 I'm not sure if I got this right - Joachim ________________________________ Von: Houghton,Andrew [mailto:houghtoa@oclc.org] Gesendet: Donnerstag, 22. Januar 2009 17:14 An: Neubert Joachim; public-esw-thes@w3.org Betreff: RE: substitute for skos:isSubjectOf Yup, I totally didn't understand what you meant. I thought http://zbw.eu/stw/descriptor/19289-4 was the bibliographic resource not the concept. However given that new information it's just a matter of switching the URI's used in the example. The rdf:about on the skos:Concept becomes http://zbw.eu/stw/descriptor/19289-4 and the rdf:about on the rdf:Description becomes the URI for the bibliographic resource and the rdf:resource on the dct:subject becomes http://zbw.eu/stw/descriptor/19289-4. Sorry for the confusion. Andy. From: Neubert Joachim [mailto:J.Neubert@zbw.eu] Sent: Thursday, January 22, 2009 11:04 AM To: Houghton,Andrew; public-esw-thes@w3.org Subject: AW: substitute for skos:isSubjectOf Hi Andy, I'm not sure if I understood exactly what you mean. In my example, http://zbw.eu/stw/descriptor/19289-4 is the URI of the skos:Concept. I do not see how I could express an inverse relation to dc:subject by means of dcterms:relation. As rdfs:seeAlso, dcterms:relation in itself appears to me too general to serve as an substitute for skos:isSubjectOf. Regards, Joachim ________________________________ Von: Houghton,Andrew [mailto:houghtoa@oclc.org] Gesendet: Donnerstag, 22. Januar 2009 16:38 An: Neubert Joachim; public-esw-thes@w3.org Betreff: RE: substitute for skos:isSubjectOf Right, I see a few mistakes in what I did, but assuming the URI for skos:Concept is http://www.econis.eu/concept/Black+market, then change the dct:subject to http://www.econis.eu/concept/Black+market. Now I think it says that the resource at http://zbw.eu/stw/descriptor/19289-4 has the subject Black market. Does that help? Andy. From: Neubert Joachim [mailto:J.Neubert@zbw.eu] Sent: Thursday, January 22, 2009 10:34 AM To: Houghton,Andrew; public-esw-thes@w3.org Subject: AW: substitute for skos:isSubjectOf Hi Andrew, dcterms:subject is defined as "The topic of the resource", and I don't want to say: The result list of the search http://www.econis.eu/...=Black+market is the subject of the skos:Concept w/ prefLabel "Black+market". Regards, Joachim ________________________________ Von: Houghton,Andrew [mailto:houghtoa@oclc.org] Gesendet: Donnerstag, 22. Januar 2009 16:22 An: public-esw-thes@w3.org Betreff: RE: substitute for skos:isSubjectOf Why couldn't you do something like: <?xml version="1.0"?> <rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#" xmlns:dct="http://purl.org/dc/terms/" xmlns:skos="http://www.w3.org/2008/05/skos#" > <skos:Concept rdf:about="http://www.w3.org/"> <skos:prefLabel>Black market</skos:prefLabel> <dct:relation> <rdf:Description rdf:about="http://zbw.eu/stw/descriptor/19289-4"> <dct:subject rdf:resource="http://www.econis/...=Black+market" /> </rdf:Description> </dct:relation> </skos:Concept> </rdf:RDF> From: public-esw-thes-request@w3.org [mailto:public-esw-thes-request@w3.org] On Behalf Of Neubert Joachim Sent: Thursday, January 22, 2009 9:57 AM To: public-esw-thes@w3.org Subject: substitute for skos:isSubjectOf The SKOS Primer recommends using dc:subject for indexing purposes. For the inverse relation - this concept "is subject of" or "indexes" that resource - it gives no recommendation. The discussion about indexing properties in SKOS (commendably summarized in http://www.w3.org/2006/07/SWD/wiki/SkosDesign/Indexing) also paid no attention to the now vanished skos:isSubjectOf. But it's definitively missing. My use case is this: On the upcoming RDFa-enriched web pages for the Standard Thesaurus Economics (maintained by the German National Library of Economics, ZBW), we will include links to library resources. This links trigger a search for the concept at hand in the subject index of the library database - e.g. http://www.econis.eu/DB=1/LNG=EN/CMD?ACT=SRCHM&IKT3=46&TRM3=Black+market searches for the concept http://zbw.eu/stw/descriptor/19289-4 with the prefLabel "Black market" and returns a result page with further links to books, articles and so on (not RDFa-enriched up to now, but this could change). To enhance the Web of Linked Data, I want to express this relationship in RDFa also. A triple http://zbw.eu/stw/descriptor/19289-4 <http://zbw.eu/stw/descriptor/19289-4> skos:isSubjectOf http://www.econis.eu/...=Black+market would have been perfect. (Of cause, the inverse relation can be expressed easily by dc:subject, but that puts the descriptor in the object position.) I understand that the missing relation may be, strictly spoken, not part of the KOS (but an application of the KOS to resources) and therefore may be outside the scope of SKOS. But nevertheless a substitute is required. It would be easy to define some custom property, e.g. zbw:indexes, with skos:Concept as domain and rdf:resource as range. But since I feel that this may be a requirement not only by our site, there should be a more common, standardized way. I did some research on this, but was out of luck: - dc:subject lacks an inverseProperty - sioc:topic dito - dcterms:references/dcterms:isReferencedBy seems to be aimed at citations and alike - ore:describes/ore:isDescribedBy is too restricted in Domain/Range (ore:ResourceMap/ore:Aggregation) - rdfs:seeAlso is too general - frbr:isSubjectOf could be fine if domain and range are sufficiently general, but I couldn't figure out the status of this one (it is described as a skos:Concept in http://sandbox.metadataregistry.org/concept/show/id/923.html, but is not part of http://purl.org/vocab/frbr/core or http://purl.org/vocab/frbr/extended) I'm convinced that terminologies - expressed in SKOS - could perfectly serve as hubs in a Web of Linked Data. To this end, it would be highly beneficial to figure out and recommend a uniform way to link from skos:Concepts to arbitrary other resources. Can we do this? Kind regards, Joachim
Received on Saturday, 24 January 2009 13:45:18 UTC