- From: Hugh Glaser <hg@ecs.soton.ac.uk>
- Date: Wed, 29 Jul 2009 12:44:14 +0100
- To: Kingsley Idehen <kidehen@openlinksw.com>
- CC: Eric Lease Morgan <eric_morgan@infomotions.com>, "public-lod@w3.org" <public-lod@w3.org>
On 29/07/2009 12:35, "Kingsley Idehen" <kidehen@openlinksw.com> wrote: > Hugh Glaser wrote: >> On 28/07/2009 14:46, "Kingsley Idehen" <kidehen@openlinksw.com> wrote: >> >> >>> Hugh Glaser wrote: >>> >>>> Good stuff. >>>> However, I don't think that Named Graphs are the answer. >>>> I get my Linked Data by resolving URIs over http. >>>> If I ask your Linked Data Space (I hope that is the right use of your >>>> terminology) for something like >>>> curl -H "Accept: application/rdf+xml" http://dbpedia.org/resource/London >>>> and follow the redirect don't I still get the non-wikipedia data with the >>>> wikipedia data? >>>> Or am I not understanding something? >>>> >>>> >>> The link chain shouldn't be broken. Named Graphs should have zero impact >>> on HTTP URIs. >>> >> That is what I thought. >> So how is the linkage data kept separate when I do URI resolution? >> Cheers >> > Hugh, > > The Linkage isn't what needs separating. > > Its when you make a data set that is 100% entity to entity links > triples (i.e., a linkset or linkbase) that needs separating (as good > practice) from the main KB. Remember, there are times when the main KB > and the source of cross links to external entities are produced by > separate parties. Thus, the linksets end up in their own Named Graphs. > Purely for organization and maintenance. This kind of partitioning > allows the use SPARUL scoped to Named Graphs when fixing triple > statement errors (e.g. owl:sameAs triples), for instance. That's great - I think you agree with me. :-) As I think when you say Named Graph you mean a different URI scheme for linkage information. Cheers Hugh > > Kingsley > > > >>> I think Alan is saying: put what is best described as a linkbase dump in >>> a separate Named Graph. Doing this shouldn't break the tapestry inherent >>> in the HTTP URIs (the data conductors). We have tons of data in >>> <http://lod.openlinksw.com> partitioned across named graphs, and none of >>> that breaks the "follow-your-nose" pattern. Remember, I am a stickler >>> for keeping the HTTP URIs of entities in full scope of user agents :-) >>> >>> The only time you might have an issue is when performing SPARQL, where >>> explicitly identifying the Named Graph in the FROM Clause may aid >>> performance (and even here this depends on the indexing in placece re, >>> the RDF DBMS insta, these days re. Virtuoso that doesn't even matter >>> since the default indexing scheme has been changed). >>> >>> Kingsley >>> >>>> Best >>>> Hugh >>>> >>>> >>>> On 28/07/2009 11:17, "Kingsley Idehen" <kidehen@openlinksw.com> wrote: >>>> >>>> Hugh Glaser wrote: >>>> >>>> >>>>> For the record ( © Alan!). >>>>> I consider it bad practice to keep the knowledge about linking in the same >>>>> KB as the substantive knowledge you are representing. >>>>> You need two KBs: one for the knowledge you are publishing, and one for >>>>> the >>>>> linkage you are working on. >>>>> These have very different provenance, maintenance patterns, etc.. >>>>> And you can include a link from URIs that you generate to the linkage KB. >>>>> >>>>> >>>>> >>>> For terminology consolidation purposes, what you call a KB is a >>>> "Linked Data Space" in my parlance :-) >>>> >>>> Yes, the partitioning suggested above is very important. Thus, you need >>>> purpose specific Linked Data Spaces (hosing many Named Graphs) if you >>>> seek to make things a little clearer to data consumers and their agents. >>>> >>>> >>>>> In fact, this would then help Alan's problem about sameAs:- he could >>>>> simply >>>>> decide not to get your view of the linkage, whereas with sameAs in the >>>>> resources he has no choice but to accept your view, and even your >>>>> predicate >>>>> when he resolves a URI or queries the SPARQL. >>>>> >>>>> And I do agree with you about minting URIs to your local stuff, including >>>>> authors; it is error-prone to try to re-use things like dbpedia for this, >>>>> on >>>>> any scale. And this is why you need to tackle the linkage problem as a >>>>> separate engineering activity. >>>>> >>>>> Best >>>>> Hugh >>>>> >>>>> (Of course I do have some software and architecture that supports separate >>>>> linkage KBs (our CRS) so I would say this, but nevertheless I think it is >>>>> the correct engineering approach, however it is done. Separation of >>>>> Concerns.) >>>>> >>>>> >>>>> >>>> Note, we've partitioned DBpedia in such a way that you now have a Graph >>>> IRI for each data set within this particular Linked Data Space. >>>> >>>> Kingsley >>>> >>>> >>>>> On 28/07/2009 02:23, "Eric Lease Morgan" <eric_morgan@infomotions.com> >>>>> wrote: >>>>> >>>>> >>>>> >>>>> On Jul 25, 2009, at 5:09 AM, Bill Roberts wrote: >>>>> >>>>> >>>>> >>>>> >>>>>> Regarding linking to external resources, what it seems you want >>>>>> to do is to identify the dc:creator of the book, hence say that >>>>>> the creator is the person whose name was Thomas More. You could >>>>>> create your own URI and if you are managing a whole bunch of data >>>>>> about books and authors, then there could be reasons to do that, >>>>>> but in general if there is a satisfactory existing URI, it is >>>>>> preferable to use it. Dbpedia seems to have become the de facto >>>>>> standard... >>>>>> >>>>>> >>>>>> >>>>> Okay, then how's this for a recipe to create rich linked data of >>>>> electronic books and authors within my own site as well as to the >>>>> outside world: >>>>> >>>>> 1. Mint URIs pointing to representations of local etexts >>>>> 2. Mint URIs pointing to representations of authors of local etexts >>>>> >>>>> 3. In resources of etexts, include owl:sameAs links to DBpedia >>>>> resources >>>>> 4. In resources of etexts, point to local URIs of authors >>>>> >>>>> 5. In resources of authors, include owl:sameAs links to DBpedia >>>>> resources >>>>> 6. In resources of authors, include owl:creatorOf links to local >>>>> etexts >>>>> >>>>> 7. For extra credit, do the same thing for subjects/keywords >>>>> >>>>> For example, the following resource descriptions: >>>>> >>>>> <!-- etext #1; points to local author and remote title --> >>>>> <rdf:RDF >>>>> xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#" >>>>> xmlns:dcterms="http://purl.org/dc/terms/" >>>>> xmlns:owl="http://www.w3.org/2002/07/owl#"> >>>>> <rdf:Description >>>>> rdf:about="http://infomotions.com/etexts/id/more-utopia-221" >>>>> owl:sameAs="http://dbpedia.org/resource/Utopia_(book)"> >>>>> <dcterms:title>Utopia</dcterms:title> >>>>> <dcterms:creator >>>>> rdf:resource="http://infomotions.com/etexts/authors/resource/thomas-more >>>>> " /> >>>>> </rdf:Description> >>>>> </rdf:RDF> >>>>> >>>>> >>>>> <!-- etext #2; points to local author and remote title --> >>>>> <rdf:RDF >>>>> xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#" >>>>> xmlns:dcterms="http://purl.org/dc/terms/" >>>>> xmlns:owl="http://www.w3.org/2002/07/owl#"> >>>>> <rdf:Description >>>>> rdf:about="http://infomotions.com/etexts/id/more-reality-404" >>>>> owl:sameAs="http://dbpedia.org/resource/Reality_(book)"> >>>>> <dcterms:title>Reality</dcterms:title> >>>>> <dcterms:creator >>>>> rdf:resource="http://infomotions.com/etexts/authors/resource/thomas-more >>>>> " /> >>>>> </rdf:Description> >>>>> </rdf:RDF> >>>>> >>>>> >>>>> <!-- author; points to local etexts and remote author --> >>>>> <rdf:RDF >>>>> xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#" >>>>> xmlns:owl="http://www.w3.org/2002/07/owl#"> >>>>> <rdf:Description >>>>> rdf:about="http://infomotions.com/etexts/authors/resource/thomas-more >>>>> " >>>>> owl:sameAs="http://dbpedia.org/resource/Thomas_More"> >>>>> <owl:creatorOf >>>>> rdf:resource="http://infomotions.com/etexts/id/more-utopia-221 >>>>> "/> >>>>> <owl:creatorOf >>>>> rdf:resource="http://infomotions.com/etexts/id/more-reality-404 >>>>> " /> >>>>> </rdf:Description> >>>>> </rdf:RDF> >>>>> >>>>> -- >>>>> Eric Lease Morgan >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>> -- >>>> >>>> >>>> Regards, >>>> >>>> Kingsley Idehen Weblog: http://www.openlinksw.com/blog/~kidehen >>>> President & CEO >>>> OpenLink Software Web: http://www.openlinksw.com >>>> >>>> >>>> >>>> >>>> >>>> >>>> >>>> >>> -- >>> >>> >>> Regards, >>> >>> Kingsley Idehen Weblog: http://www.openlinksw.com/blog/~kidehen >>> President & CEO >>> OpenLink Software Web: http://www.openlinksw.com >>> >>> >>> >>> >>> >>> >> >> >> > > > -- > > > Regards, > > Kingsley Idehen Weblog: http://www.openlinksw.com/blog/~kidehen > President & CEO > OpenLink Software Web: http://www.openlinksw.com > > > > >
Received on Wednesday, 29 July 2009 11:45:14 UTC