Hi Rinke & thanks for your report, besides Kinglsey's comments, some notes from my side
On Tue, Sep 8, 2015 at 2:19 PM, Rinke Hoekstra <rinke.hoeks...@vu.nl> wrote: > Hi all, > > At the VU University Amsterdam, we're teaching various courses on the > Semantic Web in which (understandably) DBPedia plays an important role. > > Over the past year's we have carefully groomed and curated a collection of > example SPARQL queries against DBPedia. These queries revolve around cities > with the "020" area code. > > To my surprise, these queries suddenly stopped working. Apparently because > there has been a new release of DBPedia in which this was changed. > > ... however... it turns out that they work sometimes but do not work at > other times... madness! ;-) > > A bit of background: > > The examples use the <http://dbpedia.org/ontology/areaCode> property > (i.e. a curated property), and assume the area codes are represented as > string literals, e.g. "020", sometimes with an erroneous language tag. > Since 2014 release (09/2014) we removed language tags from all xsd:strings and introduced rdf:langString according to RDF 1.1 so dbpedia-owl:areaCode (not dbp:langCode) has no language tags in our dumps for 1+ year. dbp:areaCode comes from the infobox extractor that provides raw (and many time inaccurate) data. The datatype of dbp:areaCode is decided on every triple instance based on the data we get by a greedy algorithm > It now seems that in some versions of the DBPedia endpoint, the property < > http://dbpedia.org/ontology/areaCode> has been demoted to < > http://dbpedia.org/property/areaCode> (i.e. a non-curated version) in > *all* data: there is no place name with a dbpedia-owl:areaCode property. > Even though the areaCode property is still defined in the ontology (at > least, that's what I see when I point my browser at it). > Nothing is denoted, we use the mappings wiki to map infobox templates to rdf, I can imagine that a template in Wikipedia possibly changed and the mapping was not adjusted accordingly. If you can provide example resources we can identify the source of the error easier and fix it. > In addition to that (and much much worse), in some versions of the DBPedia > endpoint, area codes are represented as *integers*, but are syntactically > still presented as the value they had as strings. E.g. the area code for > Amsterdam is represented as 020 rather than "020". Needless to say that 020 > as an integer is equal to 20.... hm.... that's not what we want Area codes > simply are not integers, they are strings, because the actual digits > matter. > as mentioned before, dbp:areaCode is not to be trusted, provided only for completion. dbpedia-owl:areaCode should be consistent all times, if you find a counter example please report it. > (also, application developers should be able to rely on relative stability > of statements that use the dbpedia ontology namespace, also, the ontology > should only change monotonously: only add classes, properties etc., don't > remove them!) > in general we agree but sometime you need to break some eggs to be able to move forward and provide more consistent data. However, I don't think we made any breaking changes to dbpedia-owl:areaCode http://mappings.dbpedia.org/index.php?title=OntologyProperty:AreaCode&action=history > > In short: > A couple of months ago, the area codes were represented as language tagged > strings using the dbpedia-owl:areaCode property > this should be before August 2014. If you found it afterwards it is probably a bug > Yesterday morning at approx. 10 am, the area codes were represented as > integers using the dbpedia-prop:areaCode property > see above comments regarding dbp > Yesterday evening, at approx. 10 pm, the area codes were represented as > language tagged strings using the dbpedia-owl:areaCode property > (Again) This should be a bug if true > > Today, at approx. 1 pm, the area codes were represented as integers using > the dbpedia-prop:areaCode property > see above comments regarding dbp > > > ... > > It would be exceedingly nice if this could be fixed. Perhaps it has > something to do with load balancing and different stores not being > synchronized properly? > > -Rinke > > PS ... area codes may be strings... but they don't have to be > language-tagged ;-) > > > > ------------------------------------------------------------------------------ > > _______________________________________________ > Dbpedia-discussion mailing list > Dbpedia-discussion@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/dbpedia-discussion > > -- Kontokostas Dimitris
------------------------------------------------------------------------------
_______________________________________________ Dbpedia-discussion mailing list Dbpedia-discussion@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/dbpedia-discussion
