On 09/25/2017 01:58 AM, Markus Freudenberg wrote: > Hi Peter, > > > I would instead say that the reason here is an incorrect mapping. I don't > > see how mapping things with a Controllo_di_autorità template was ever > > sufficient or current. > > Semantics. As the history of the wikipage suggests, this is based on a > single input without any revision. Therefore, an incorrect mapping is not > surprising based on that. Mappings in these "secondary" languages (DBpedia > languages between 1M and 3M resources) are not strong, this is why we are > going for (semi-)automation in the future.
So then I guess the takeaway is not to use mappings from these languages in 2016-10 DBpedia. That's rather unfortunate, but I guess that the linguistic information there is usable. Is there a list of the primary DBpedia languages, or is English the only primary DBpedia language? > > If I found three different incorrect mappings in three different languages just by looking at two nodes, > > how many incorrect mappings are there in this DBpedia? > > Yet, I don't see any effort on your side to correct this (at least for > now). Since I haven't taken the snapshot of mappings for the coming > release, changes would be included still (This is what 'community > effort' means regarding DBpedia). I was not aware that there was any way to fix the problem. How can I do something to remove the incorrect information from, for example, downloads.dbpedia.org/2016-10/core-i18n/it/instance_types_it.ttl.bz2 I know that I could go to the mapping wiki and change the mappings and this could affect DBpedia in future (except that will this make any difference for the new mappings?), but I don't really have the expertise to do so in many cases. For example, to fix http://mappings.dbpedia.org/index.php/Mapping_it:Controllo_di_autorit%C3%A0 I need to understand the Wikipedia template at https://it.wikipedia.org/wiki/Template:Controllo_di_autorit%C3%A0 but that page is in Italian and my knowledge of Italian is nowhere near adequate to understand it. I could use Google translate to get a rough gloss in English (and I did) so I think I know more or less what is going on, but I'm not certain that my understanding is adequate to fix the problem. I did look at the history of the page and noticed that the last (only?) edit was by Ninniuz who was very active in late 2013 and early 2014, but doesn't appear to have been active at all since. I just left a message on that user's talk page. > > How can I get a reasonably large amount of information from DBpedia 2016-10 that isn't likely > > to have these kinds of errors? > > By editing/correcting the mappings. Again, how is that going to change DBpedia 2016-10? > > It sticking to the English mappings necessary? > > Is this even adequate? > > I'm not sure what you mean by sticking to English mappings? The errors that I noticed are not coming from DBpedia mappings for English Wikipedia. Maybe if I stick to the mapping facts coming from English DBpedia I won't get so many errors. > > To get less incorrect information in DBpedia in the future is likely going > > to > > require quite a bit of work. > > You are right about this. But this was always the case and we are > constantly working in that direction, hoping to get help any party > interested. > > I don't think that just switching to a different kind of mapping is going to help much. > > No but it's the basis for many depending improvements (Better UI, automatic validation, etc) > > > It should have been easy to determine that template mapping here was incorrect. > > The approach of DBpedia was to rely on the mapping community to take care > of auch mistakes. We then apply mappings and ontology to a given set of > source data, making sure that software and workflow for extracting the > data is up to date. This approach is about to shift as you glean see from > the ongoing discussion about the ontology and mappings. > Since you already spotted some mappings mistakes, I would kindly ask you > to contribute and edit the mappings to your contentment. As I stated above, I don't have the linguistic expertise to completely understand what is going on in most languages, so I am leery about changing anything there. In any case, I don't think that I can do anything to change 2016-10 DBpedia. > Best, > > Markus Freudenberg > Release Manager, DBpedia peter > On Fri, Sep 22, 2017 at 8:29 PM, Peter F. Patel-Schneider <pfpschnei...@gmail.com> wrote: > > I would instead say that the reason here is an incorrect mapping. I > don't > see how mapping things with a Controllo_di_autorità template was ever > sufficient or current. > > But what can one do with DBpedia 2016-10? If I found three different > incorrect mappings in three different languages just by looking at two nodes, > how many incorrect mappings are there in this DBpedia? How can I get a > reasonably large amount of information from DBpedia 2016-10 that isn't likely > to have these kinds of errors? It sticking to the English mappings necessary? > Is this even adequate? > > To get less incorrect information in DBpedia in the future is likely going to > require quite a bit of work. I don't think that just switching to a different > kind of mapping is going to help much - it should have been easy to determine > that the template mapping here was incorrect. > > peter > > > On 09/22/2017 07:36 AM, Markus Freudenberg wrote: > > The reason for this are insuffizient/outdated mappings. > > > > For example: > > <http://it.dbpedia.org/resource/Arti_visive> > > <http://www.w3.org/1999/02/22-rdf-syntax-ns#type> > > <http://dbpedia.org/ontology/Person> . > > (should be owl:Thing or missing all together) > > > > The wikipedia template used for https://it.wikipedia.org/wiki/Arti_visive is > > 'Controllo_di_autorità' > > The mappings wiki states: 'Controllo_di_autorità' is to be mapped to dbo:Person: > > http://mappings.dbpedia.org/index.php/Mapping_it:Controllo_di_autorità > > <http://mappings.dbpedia.org/index.php/Mapping_it:Controllo_di_autorit%C3%A0> > > > > To resolve this, change these template mappings. > > > > Way forward: > > > > 1. switch to RML mappings (in October - easier to validate) > > 2. automate the creation of template (stumps), based on similar templates in > > other languages > > 3. DBpedia fusion and type generation heuristics > > > > Sorry, don't have time to expand on this today. > > > > Best, > > > > Markus Freudenberg > > > > Release Manager, DBpedia <http://wiki.dbpedia.org> > > > > On Thu, Sep 21, 2017 at 6:03 PM, Peter F. Patel-Schneider > > <pfpschnei...@gmail.com <mailto:pfpschnei...@gmail.com>> wrote: > > > > When I created a KB out of what I thought should be the high-quality > > information in the canonicalized 2016-10 dataset of DBpedia I noticed that > > there are some systematic errors in the types of nodes. For example, Tree, > > http://wikidata.dbpedia.org/resource/Q10884 > > <http://wikidata.dbpedia.org/resource/Q10884>, is an instance of > both > > http://dbpedia.org/ontology/Agent <http://dbpedia.org/ontology/Agent> and > > http://dbpedia.org/ontology/WrittenWork > > <http://dbpedia.org/ontology/WrittenWork>, as well as a lot of other incorrect > > types. Vegetable, http://wikidata.dbpedia.org/resource/Q11004 > > <http://wikidata.dbpedia.org/resource/Q11004>, has similar > > problems > > > > I traced these errors back to the following files: > > instance_types_wkd_uris_eo.ttl:<http://wikidata.dbpedia.org/resource/Q10884 <http://wikidata.dbpedia.org/resource/Q10884>> > > <http://www.w3.org/1999/02/22-rdf-syntax-ns#type > > <http://www.w3.org/1999/02/22-rdf-syntax-ns#type>> > > <http://dbpedia.org/ontology/Bird <http://dbpedia.org/ontology/Bird>> . > > instance_types_wkd_uris_it.ttl:<http://wikidata.dbpedia.org/resource/Q10884 <http://wikidata.dbpedia.org/resource/Q10884>> > > <http://www.w3.org/1999/02/22-rdf-syntax-ns#type > > <http://www.w3.org/1999/02/22-rdf-syntax-ns#type>> > > <http://dbpedia.org/ontology/Person <http://dbpedia.org/ontology/Person>> . > > instance_types_wkd_uris_ru.ttl:<http://wikidata.dbpedia.org/resource/Q10884 <http://wikidata.dbpedia.org/resource/Q10884>> > > <http://www.w3.org/1999/02/22-rdf-syntax-ns#type > > <http://www.w3.org/1999/02/22-rdf-syntax-ns#type>> > > <http://dbpedia.org/ontology/Book <http://dbpedia.org/ontology/Book>> . > > instance_types_wkd_uris_eo.ttl:<http://wikidata.dbpedia.org/resource/Q11004 <http://wikidata.dbpedia.org/resource/Q11004>> > > <http://www.w3.org/1999/02/22-rdf-syntax-ns#type > > <http://www.w3.org/1999/02/22-rdf-syntax-ns#type>> > > <http://dbpedia.org/ontology/Bird <http://dbpedia.org/ontology/Bird>> . > > instance_types_wkd_uris_it.ttl:<http://wikidata.dbpedia.org/resource/Q11004 <http://wikidata.dbpedia.org/resource/Q11004>> > > <http://www.w3.org/1999/02/22-rdf-syntax-ns#type > > <http://www.w3.org/1999/02/22-rdf-syntax-ns#type>> > > <http://dbpedia.org/ontology/Person <http://dbpedia.org/ontology/Person>> . > > instance_types_wkd_uris_ru.ttl:<http://wikidata.dbpedia.org/resource/Q11004 <http://wikidata.dbpedia.org/resource/Q11004>> > > <http://www.w3.org/1999/02/22-rdf-syntax-ns#type > > <http://www.w3.org/1999/02/22-rdf-syntax-ns#type>> > > <http://dbpedia.org/ontology/Book <http://dbpedia.org/ontology/Book>> . > > > > I then looked at instance_types_it.ttl and noticed that there were a lot of > > incorrect instances of http://dbpedia.org/ontology/Person > > <http://dbpedia.org/ontology/Person>. From looking at > > the first few lines of the file with this type it appears that a > large > > majority of them are incorrect. It thus appears that something has gone > > very wrong in the extraction of information for Italian DBpedia. Similarly > > it appears that something has gone very wrong in the extraction of > > information for [Spanish] DBpedia. I can't make sense of the analogous > > file in Russian DBpedia, but it appears to have far too many instances of > > Book indicating that there is something very wrong there as well. > > > > The large number of errors that I have uncovered means that I can't count on > > information from these parts of DBpedia. That's regrettable, as I would > > like to include as much information as possible. But what is really > > problematic is that now I don't see how I can count on any DBpedia > > information. > > > > What is the way forward here? Are there some parts of DBpedia that are > > known not to have these sorts of systematic problems. > > > > > > Peter F. Patel-Schneider ------------------------------------------------------------------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, Slashdot.org! http://sdm.link/slashdot _______________________________________________ DBpedia-discussion mailing list DBpedia-discussion@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/dbpedia-discussion
