- From: Kingsley Idehen <kidehen@openlinksw.com>
- Date: Sun, 24 May 2009 22:43:06 -0400
- To: Hugh Glaser <hg@ecs.soton.ac.uk>
- CC: "richard.hancock@3kbo.com" <richard.hancock@3kbo.com>, "public-lod@w3.org" <public-lod@w3.org>
Hugh Glaser wrote: > On 25/05/2009 00:40, "Kingsley Idehen" <kidehen@openlinksw.com> wrote: > > >> Kingsley Idehen wrote: >> >>> Hugh Glaser wrote: >>> >>>> An interesting question - should be a classic for Linked Data. >>>> As I understand it, your primary problem for New Zealand is that >>>> there is no reliable information about the administrative geography. >>>> It doesn't help for New Zealand (!), but for the UK you could use the >>>> Ordnance Survey's, which can be found in RDF via >>>> http://osdir.com/ml/web.semantic.linking-open-data/2008-05/msg00011.html >>>> We have made it linked data at os.rkbexplorer.com, with a SPARQL >>>> endpoint if you want it. >>>> (If people are doing this sort of thing (if you thought of doing some >>>> of it yourself), then reusing the OS ontology might be a good idea.) >>>> >>>> For the UK, it would be a case of jumping (follow-your-nose) >>>> backwards and forwards between the os and dbpedia, I suggest, using >>>> the os to find the inclusion and dbpedia to find the knowledge. >>>> This afternoon I did a bunch of links between them (the bigger areas >>>> and towns), to let you have a go if you like, and these can be found >>>> at our os coreference service (crs) at >>>> http://os.rkbexplorer.com/crs/ >>>> >>>> By the way, if you want an easy way of finding out what URIs might >>>> tell you what is known about a particular place (or anything), you >>>> could try our sameAs service at >>>> http://www.rkbexplorer.com/sameAs/?uri= >>>> So New Zealand would be >>>> http://www.rkbexplorer.com/sameAs/?uri=http://dbpedia.org/resource/New_Zeala >>>> nd >>>> >>>> And Wellington >>>> >>>> > http://www.rkbexplorer.com/sameAs/?uri=http://dbpedia.org/resource/Wellingto>>> > n > >>>> and you can follow-your-nose from there. >>>> >>>> >>> Hugh, >>> >>> Assuming you sent this response before mine re. >>> http://lod.openlinksw.com . How does the following not resolve this >>> problem: >>> >>> 1. Go to http://lod.openlinksw.com >>> 2. Enter pattern: New Zealand >>> 3. Use "Type" and/or "Property" to locate what you want >>> 4. Once found, click on "statistics" link >>> >>> Kingsley >>> >>> >> I should have added: >> >> 5. look at: http://lod.openlinksw.com/fct/facet.vsp?cmd=load&fsq_id=3011 >> (do use the "retry" button to maximize results) >> >> > Hi Kingsley, > Actually, I think I had seen your response. > In answer to your question, it is different most obviously it uses > different sources. > And of course in the Linked Data world, using different sources will deliver > different results; it is never the case that one source has all the > knowledge. > I am not assuming once source. Of course not. I am assuming a possible beachhead :-) > More fundamentally, the process you present seems to be based on strings. > Of course not, it is associating entities in a particular linked data space with a full text search patter. Each of the entities have de-referencable URIs etc.. You find the specific entity in question by navigating entity type and/or properties. > I think the whole point about Linked Data is that we do not rely on the > string matching, but use URIs. > See my comments above. The whole point of Linked Data should be to demonstrate how it embraces and extends the Google full text search realm which is autistic to entities, entity types, and entity properties re. disambiguation of queries (or as they call them: searches). > That is exactly what Richard was trying to do; having found a URI such as > http://dbpedia.org/resource/New_Zealand that he is satisfied captures the > concept with which he is concerned, he now wants to explore what is known > about it in the Linked Data world, without going back to the text world. > Again, I don't think I am sending him back to the full text pattern world. I am saying: 1. Enter a patter: New Zealand (as you would re. Google, Yahoo! etc..) 2. When presented with hits (which are really Entities with URIs plus excerpts from associated literal object values) filter further by Entity Type or Entity Property > And I donıt think he wanted to do any clicking he wanted to just script it > all up in a reliable Linked Data sort of way. > Lets assume he didn't want to click anything, what do you think the purpose of the "URI Lookup by Label" and "URI Lookup" tabs are for then? They are for entering patterns that are associated with Entity Labels or actual URIs. The instance at: <http://lod.openlinksw.com> is but one data space on the vast Web of Linked Data. It's a linked data junction box with lots of de-referencable URIs that can take you to many places on the Web or conduct data via many pathways on the Web. I don't understand why you find my responses fundamentally incongruent with the very essence of Linked Data. We keep on going round the same loop in different ways. Kingsley > Best > Hugh > > PS By the way, I tried putting his New Zealand and Wellington URIs into the > URI Lookup search box at http://lod.openlinksw.com/fct/ and canıt seem to > get anything as it always only offers stuff with them extended by #... > presumably I am doing something wrong? > PPS I donıt see a ³retry² button. > > > -- Regards, Kingsley Idehen Weblog: http://www.openlinksw.com/blog/~kidehen President & CEO OpenLink Software Web: http://www.openlinksw.com
Received on Monday, 25 May 2009 02:43:43 UTC