- From: Sue Ellen Wright <sellenwright@gmail.com>
- Date: Sat, 20 May 2006 10:31:24 -0400
- To: "Mark van Assem" <mark@cs.vu.nl>, "Alan Melby" <melbyak@yahoo.com>
- Cc: "Paul Hermans" <paul.hermans@amplexor.com>, public-esw-thes@w3.org
- Message-ID: <e35499310605200731s5fa9fbeclcda49bd5aefd1190@mail.gmail.com>
Hi, Mark, I just see that the earlier response I sent to Paul was just to him, so I'm adding it here from the standpoint of terminological markup. I can't speak for whether there is a skos mechanism, but in terminology markup it becomes combersome: we provide a target marker or termid for each term in the system, and then other terms can use a pointing mechanism to target the equivalent term. Usually it's enough to have all the terms together in a concept entry, but your question zeros in on specific term to term matches, which requires additional internal markup. Last year we talked about creating a cross-walk between the TMF/LMF terminology and lexicography world (both rdf formats) and SKOS. We did some preliminary mapping, but we haven't had the bandwidth to go beyond that, and we're still finishing the LMF, so it's a little premature. Paul's concern is a typical one in that community of practice because we may indeed want to state that we have a broad overall concept for rest/repose, etc., and there may be languages where there is only one term to cover the whole concept. It just happens that between these two intimately related languages (doesn't get closer than this unless we move out to Friesian, which with all due respect to my fine Friesian friends, doesn't come up that frequently in standard thesauri) we have internal links as well as broader conceptual ones. They really reflect differences in register and cultural nuance, which translators and multilingual communicators find extremely useful. For instance, since I speak German and English, but only play around with Dutch, I might not know precisely that A does indeed = B in the relationship between the two languages. In fact, these subtle differences abound between Dutch and English terms and concepts, which makes establishing equivalence between the two much more difficult than one might suspect. [Of course Paul and Mark know that -- I'm lecturing to the wall here.] I guess the question that arises, however, is how much can we load SKOS (it's supposed to be simple after all!) down with fine semantic and linguistic detail. Right now the terminologists are of a mind that we should use our own solutions for full-blown terminology and lexicography representation and map our concept system elements to SKOS without trying to achieve a lossless round trip for all terminological and lexicographical distinctions. In a concrete case, one might build the crosswalk as a bridge between a SKOS resource and a lex/term resource, with the SKOS representation providing detailed semantic linkages and maybe an entry to the OWL world, and the TMF/LMF side providing more linguistic detail. Of course, that's all sort of visionary at the moment because, as noted, we haven't had time to do it in real life! Bye for now Sue Ellen On 5/20/06, Mark van Assem <mark@cs.vu.nl> wrote: > > > Hi Paul, > > Conceptually speaking, the idea in SKOS is that each pref/altLabel is in > some sense a correct lexicalization of the concept (in the scope of the > particular vocab), so no relations between the labels would have to be > added. > > Technically speaking it is even impossible to state relations between > terms. It is only possible to state relations between concepts. That is > one of the drawbacks of the approach SKOS takes (see also [1]). > > A possible way out is to suppose that the rusten/repose pair actually > represents a different concept than the serenity/sereniteit pair and > create two separate skos:Concepts for them, and assume that > pref/altLabels of different languages in one concept are already correct > translations of each other. > > Cheers, > Mark. > > [1]http://lists.w3.org/Archives/Public/public-esw-thes/2006May/0009 > > Paul Hermans wrote: > > Dear, > > > > Example: > > <skos:Concept rdf:about="" > > > <skos:prefLabel xml:lang="en">peace</skos:prefLabel> > > <skos:prefLabel xml:lang="nl">rust</skos:prefLabel> > > <skos:altLabel xml:lang="en">repose</skos:altLabel> > > <skos:altLabel xml:lang="en">serenity</skos:altLabel> > > <skos:altLabel xml:lang="nl">rusten</skos:altLabel> > > <skos:altLabel xml:lang="nl">sereniteit</skos:altLabel> > > </skosConcept> > > > > Is there a way in SKOS core to indicate that "rusten" is the translation > of "repose" and "sereniteit" of "serenity"; > > that these altLabels belong together? > > > > > > Regards, > > > > > > Paul > > > > > > -- Sue Ellen Wright Institute for Applied Linguistics Kent State University Kent OH 44242 USA sellenwright@gmail.com swright@kent.edu sewright@neo.rr.com
Received on Saturday, 20 May 2006 14:31:35 UTC