- From: Hugh Glaser <hg@ecs.soton.ac.uk>
- Date: Sat, 5 Jul 2008 20:52:28 +0100
- To: "public-lod@w3.org" <public-lod@w3.org>
- CC: Bernard Vatant <bernard.vatant@mondeca.com>
Hi Giovanni, Thanks. Yes, we have been worrying about this for a while, and been trying to tease apart some of the requirements. On 05/07/2008 19:10, "Giovanni Tummarello" <giovanni.tummarello@deri.org> wrote: > Hi Bernard, there were talks on LOD some time ago (now migrating the > dicussion there) from Hugh Glaser if i am not wrong about using > something different from sameAs, i dont recall precisely. http://eprints.ecs.soton.ac.uk/15614/ There is another report at http://eprints.ecs.soton.ac.uk/15765/ > > Hugh was it ever formally defined some alternative to sameAs? i think > if there is something witha clear use and consensus we should add it > in the publishing linked data tutorials. There is a little ontology at http://www.rkbexplorer.com/ontologies/coref It used to be bigger (!), but we found we didn't use the rest of it (yet), so we got rid of those bits for the moment. > > but on the other hand.. i think one always has to cautiously and > defensively interpret statements that are found on the semantic web, > so with respect to this, automaticlaly turning "sameAs" into something > less powerful or directional before doing one's reasoning seems > nevertheless a good idea. We have a feeling that people who want to publish knowledge about equivalence-style relationships should be able to do so, and that the semantics might be "weak" in some sense (because they are based on context), so that they are open to interpretation by the consumer of the knowledge (maybe by considering as sameAs or perhaps seeAlso or ignoring). Perhaps what we are trying to do is empower the consumer of knowledge (by which I mean agent) to make decisions about what to do with it (as well as where to get it). And perhaps this is why it relates to deprecation; in one meaning of deprecation, you are not forcing me to do anything - you are simply telling me what you would like me to do. In this sense the structure is similar to your (Giovanni's) description of how sameAs might work in practice (at a Web scale); although it is not actually directional, I need to find out what URIs are the sameAs the one in my hand, and go to services/KBs that tell me. I am very much in agreement with your descriptions of the interactions and your "pragmatic interpretation". Caveat emptor, as they used to say in Italy. So we envision the provision of such services (we call ours CRSes); the slight difference being that we will give you sameAs if you ask, but will also give you the "weak"er version, and think you will probably prefer it, and then make your own decisions about what to do about inference. So we have very much tried to separate out the publishing from the consuming. A couple of things in addition. There are times when I might use a raw owl:sameAs to publish my knowledge: the way I think of it is that this is when I find another URI and realise I would/might have used that URI instead of mine, had I known of its existence. We will need to be able to represent context of the equivalence KBs (such as CRS), so consumers can make informed choices. This is likely to be little different from the same requirement for any KB, so we await the ability to publish knowledge of context, trust, provenance,... in the general case! Best Hugh > > Giovanni > > On Fri, Jul 4, 2008 at 10:17 AM, Bernard Vatant > <bernard.vatant@mondeca.com> wrote: >> Hi Giovanni >> >> Interesting view - not answering the initial question, though - except to >> confirm that OWL does not provide any out-of-the-box answer ... which is no >> news. :-) . >> That said, having a non-symmetrical interpretation of 'owl:sameAs' is a bit >> of a hack. Default more accurate vocabulary, owl:sameAs is used for about >> any kind of "referent similarity" (we've discussed that at length many times >> in this and other forums). In the case you explain the property to use (and >> which is indeed lacking in the standard vocabulary) should be something like >> ":compatibleWith" ... >> >> a:foo :compatibleWith b:bar >> >> ... meaning that if you can add elements of description of :bar to those of >> :foo. Which is not necessarily symmetrical. For example DBpedia has indeed >> one "Berlin" whereas Geonames has two (the administrative subdivision and >> the populated place). So we could write. >> >> <http://dbpedia.org/resource/Berlin> :compatibleWith >> <http://sws.geonames.org/2950159/> >> <http://dbpedia.org/resource/Berlin> :compatibleWith >> <http://sws.geonames.org/2950157/> >> >> Actually what we need is a namespace and vocabulary for all those flavors of >> URI similarity and equivalence to be used on the Web, diffferent from OWL >> and RDFS namespace. >> >> Bernard >> >> Giovanni Tummarello a écrit : >>>> >>>> "http:..." or something equivalent, not a reference <http:...>, since in >>>> the >>>> latter case you're talking about what the URI names, not the URI itself, >>>> and >>>> things can have more than one name - some of which might be deprecated, >>>> and >>>> >>> >>> Some thoughs: >>> >>> Pragmatically speaking if there is one thing that "linked data" buys >>> us is new URIs for the same thing others might be talking about, but >>> specifically minted within "one's context", that is the name space of >>> the site who's hosting the RDF >>> >>> so http://dbpedia.org/resource/Berlin is really not "berlin" in >>> general but obviously "something that they have in their DB that has >>> label "berlin" and bla bla >>> >>> then there is http://geonames/123456 that might also be berlin. If >>> there is an agreement by the dbpedia guys that that entity is >>> equivalent for DBPEDIA PURPOSES to their entities then the dbpedia >>> people put a "sameAs" between the two. So "sameAs" on the web of >>> linked data is always a "directed sameAs" >>> >>> This pragmatic interpretation should be "we people at dbpedia believe >>> that it might be useful for you as a robot to also go collect >>> information from this other source as we find it generally compatible >>> with the information we provide here" >>> >>> given this interpretation of "linked data URIs" (something i >>> previously called URI/URLs and i still dont find a better term for it) >>> then i believe its perfectly valid to state things about the URI >>> itself, since they always must be interpreted within the context where >>> they were taken. >>> >>> i know, OWL doesnt account for this, that is, it will not make >>> distinctions between "context only valid statements" and not.. but i >>> see no alternative to deal with this at preprocessing level, e.g. when >>> you crawler picks up linked data information, you should look for said >>> context onyl valid statments before you smush all together with the >>> sameAs >>> >>> Giovanni >>> >>> >>> No virus found in this incoming message. >>> Checked by AVG - http://www.avg.com Version: 8.0.135 / Virus Database: >>> 270.4.4/1532 - Release Date: 03/07/2008 08:32 >>> >>> >>> >>> >> >> -- >> >> *Bernard Vatant >> *Knowledge Engineering >> ---------------------------------------------------- >> *Mondeca** >> *3, cité Nollez 75018 Paris France >> Web: www.mondeca.com <http://www.mondeca.com> >> ---------------------------------------------------- >> Tel: +33 (0) 971 488 459 >> Mail: bernard.vatant@mondeca.com <mailto:bernard.vatant@mondeca.com> >> Blog: Leçons de Choses <http://mondeca.wordpress.com/> >> >> >> >
Received on Saturday, 5 July 2008 19:55:08 UTC