Speaking of wrong mappings, do the algorithms used to generate the Airpedia
class mappings have any concept of classes which are (or should be)
disjoint with each other? I was looking at the distribution of the number
of classes assigned to entities and was curious what classes were assigned
to the entities with the most classes. Naturally the very first one I
picked to look at was rather strange looking.
$ zgrep Zosimas airpedia-classes-en.nt.gz
<http://dbpedia.org/resource/Zosimas_of_Palestine> <
http://airpedia.org/ontology/type_with_conf#10> <
http://dbpedia.org/ontology/Eukaryote> .
<http://dbpedia.org/resource/Zosimas_of_Palestine> <
http://airpedia.org/ontology/type_with_conf#10> <
http://dbpedia.org/ontology/FloweringPlant> .
<http://dbpedia.org/resource/Zosimas_of_Palestine> <
http://airpedia.org/ontology/type_with_conf#10> <
http://dbpedia.org/ontology/Plant> .
<http://dbpedia.org/resource/Zosimas_of_Palestine> <
http://airpedia.org/ontology/type_with_conf#10> <
http://dbpedia.org/ontology/Saint> .
<http://dbpedia.org/resource/Zosimas_of_Palestine> <
http://airpedia.org/ontology/type_with_conf#10> <
http://dbpedia.org/ontology/Agent> .
<http://dbpedia.org/resource/Zosimas_of_Palestine> <
http://airpedia.org/ontology/type_with_conf#10> <
http://dbpedia.org/ontology/Species> .
<http://dbpedia.org/resource/Zosimas_of_Palestine> <
http://airpedia.org/ontology/type_with_conf#10> <
http://dbpedia.org/ontology/Person> .
<http://dbpedia.org/resource/Zosimas_of_Palestine> <
http://airpedia.org/ontology/type_with_conf#10> <
http://dbpedia.org/ontology/Cleric> .
Looking at the Wikipedia article, I'm not seeing where "flowering plant" is
coming from, but regardless, it should probably recognize Flowering Plant
(and its parents) as being disjoint from Person.
Has anyone compared the inferred types against the types assigned in
Freebase? That'll be my next project if no one else has already done it.
Tom
On Fri, Jun 14, 2013 at 6:37 PM, Alessio Palmero Aprosio <apro...@fbk.eu>wrote:
> *
>
> Dear DBpedians,
>
> we are the team of Airpedia project [1], which aims to enhance the
> classes/properties coverage of DBpedia over Wikipedia using machine
> learning techniques.
>
> We read about the “mapping sprint”, therefore we want to bring to your
> attention the resource we are producing concerning DBpedia. We think that
> it can help the community to speed up the mapping process.
>
> Wrong mappings
>
> The basic idea of our approach is the use of DBpedia resource as training
> data. For this reason, we have to be sure that the mappings are correct. We
> then implement a cross-language validation to discover wrong mappings. We
> found out some obvious errors, that we think may be correct before the
> release of DBpedia 3.9. See attachment for the list of these mappings.
>
> Automatic class mappings
>
> In a paper accepted to I-KNOW conference [2], we present a resource
> obtained by automatically mapping Wikipedia templates in 25 languages.
> Our approach can replicate the human mappings with high reliability, and
> producing an additional set of mappings not included in the original
> DBpedia. The resource can be downloaded from the resource section [3] of
> the Airpedia website and consists of CSV files with two columns: Wikipedia
> infobox name and DBpedia class.
>
> Automatic properties mappings
>
> In a second paper submitted to ISWC conference [4], we focus on the
> problem of automatically mapping infobox attributes to properties into the
> DBpedia ontology for extending the coverage of the existing localised
> versions or building from scratch versions for languages not covered in the
> current version. We report results comparable to the ones obtained by a
> human annotator in term of precision, but our approach leads to a
> significant improvement in recall and speed. Specifically, we mapped
> 45,978 Wikipedia infobox attributes to DBpedia properties in 14 different
> languages for which mappings were not available yet. Again, it can be
> downloaded from the resource section [3] of the Airpedia website and
> consists of CSV files with two columns: Wikipedia infobox attribute name
> and DBpedia property.
>
> Enhanced coverage of DBpedia over classes in 31 languages
>
> Following the work already presented at ESWC conference [5], we enhance
> the coverage of DBpedia over pages devoid of infobox. The resource
> contains 10M computed entity types. It is available in RDF format and can
> be downloaded in the resource section [3] of our website.
>
> Integration in Italian DBpedia
>
> The Italian DBpedia team has been the firts adopter of our dataset. Next
> week a new version of the SPARQL endpoint containing our statements will be
> released. Stay tuned!
>
> Cheers!
>
> Alessio
>
>
> [1] http://www.airpedia.org
>
> [2] http://i-know.tugraz.at/
>
> [3] http://www.airpedia.org/download/
>
> [4] http://iswc2013.semanticweb.org/
>
> [5] http://2013.eswc-conferences.org/
>
> *
>
>
> ------------------------------------------------------------------------------
> This SF.net email is sponsored by Windows:
>
> Build for Windows Store.
>
> http://p.sf.net/sfu/windows-dev2dev
> _______________________________________________
> Dbpedia-discussion mailing list
> Dbpedia-discussion@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/dbpedia-discussion
>
>
------------------------------------------------------------------------------
This SF.net email is sponsored by Windows:
Build for Windows Store.
http://p.sf.net/sfu/windows-dev2dev
_______________________________________________
Dbpedia-discussion mailing list
Dbpedia-discussion@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dbpedia-discussion
