- From: J�rn Hees <j_hees@cs.uni-kl.de>
- Date: Tue, 9 Aug 2011 12:26:51 +0200
- To: public-lod@w3.org
- Cc: dbpedia-discussion@lists.sourceforge.net
Hi, see the "Limitations on browseable data" section on http://www.w3.org/DesignIssues/LinkedData . That section pretty much covers your 2nd remark, the one afterwards your 1st one. But yes, i also consider DBpedia buggy in this sense (hence the crossposting): I guess that many newcomers won't even notice that they might not have gotten any sensible triples at all. It seems that internally the URI dereferencing results are capped at a 2001 triples max. At the same time the incoming (/ reverse) links are given first, then the outgoing links. For example, if you dereference http://dbpedia.org/resource/United_States you'll get 2001 incoming links (?s ?p dbpedia:United_States), but not even an rdfs:label: http://dbpedia.org/data/United_States.ntriples . I also guess it would be better to construct the given document first from the outgoing triples, maybe preferring the ontology mapped triples, and then incoming links up to a 2000 triples limit (if necessary to limit bandwidth). That would fit the description in the above mentioned section way better than the current implementation. Best regards, J�rn On 8. Aug. 2011, at 17:15, Basil Ell wrote: > Hi, > > I wonder about the limit of triples when accessing DBpedia URIs: > > $ rapper -c "http://dbpedia.org/resource/Netherlands" > rapper: Parsing URI http://dbpedia.org/resource/Netherlands with parser rdfxml > rapper: Parsing returned 2001 triples > > When I access that URI by browser I receive the complete data, this means that machines are underprivileged > whereas they are the ones that are capable of processing the amount of data instead of human users. > > Wouldn't it be nice to: > 1) whenever such a limit is applied to return a triple that states that a limit has been applied, > then the machine knows that it does not know everything there is to know and > 2) to include triples based on their expected relevance? For example the rdfs:label is generally of interest. > > Best regards, > Basil Ell
Received on Tuesday, 9 August 2011 10:27:40 UTC