search

dice-group / AGDISTIS

AGDISTIS - Agnostic Named Entity Disambiguation
http://aksw.org/Projects/AGDISTIS.html
GNU Affero General Public License v3.0
141 stars 37 forks source link

German DBpedia 2016-10 #47

Closed RicardoUsbeck closed 6 years ago

RicardoUsbeck commented 7 years ago

Using the following sentence: <entity>Angela Merkel</entity> was in <entity>Germany</entity>

AGDISTIS returns the following in FOX:
scms:means http://de.dbpedia.org/resource/Angela_Merici ;
scms:source source:fox ;
ann:body "Angela Merkel"^^xsd:string
scms:means http://de.dbpedia.org/resource/Germany_Schulz ;
scms:source source:fox ;
ann:body "Germany"^^xsd:string

Please fix that and write a unit test for it.

RicardoUsbeck commented 6 years ago

With the latest version the outcome is:

curl --data-urlencode "text='<entity>Angela Merkel</entity> was in <entity>Germany</entity>'" -d type='agdistis' http://localhost:8080/AGDISTIS

[{"disambiguatedURL":"http:\/\/de.dbpedia.org\/resource\/Angela_Merkel","offset":13,"namedEntity":"Angela Merkel","start":1},{"disambiguatedURL":"http:\/\/de.dbpedia.org\/resource\/Germany_(Pennsylvania)","offset":7,"namedEntity":"Germany","start":22}]
DiegoMoussallem commented 6 years ago

Question: Why should AGDISTIS be able to disambiguate "Germany" in the German version as the correct name is "Deutschland"?

Bug: AGDISTIS is not able to disambiguate "Germany" because "Germany" does not appear as a country neither in German DBpedia nor Wikipedia. It is a disambiguation page in both which actually points to many resources but none of them is the country. See:

> <http://de.dbpedia.org/resource/Germany http://dbpedia.org/ontology/wikiPageDisambiguates http://de.dbpedia.org/resource/Germany_(Mondkrater)>
> <http://de.dbpedia.org/resource/Germany http://dbpedia.org/ontology/wikiPageDisambiguates http://de.dbpedia.org/resource/Italo_Reno_&_Germany>
> <http://de.dbpedia.org/resource/Germany http://dbpedia.org/ontology/wikiPageDisambiguates http://de.dbpedia.org/resource/Germany_(Georgia)>
> <http://de.dbpedia.org/resource/Germany http://dbpedia.org/ontology/wikiPageDisambiguates http://de.dbpedia.org/resource/Germany_(Indiana)>
> <http://de.dbpedia.org/resource/Germany http://dbpedia.org/ontology/wikiPageDisambiguates http://de.dbpedia.org/resource/Germany_(Pennsylvania)>
> <http://de.dbpedia.org/resource/Germany http://dbpedia.org/ontology/wikiPageDisambiguates http://de.dbpedia.org/resource/Germany_(Texas)>
> <http://de.dbpedia.org/resource/Germany http://dbpedia.org/ontology/wikiPageDisambiguates http://de.dbpedia.org/resource/Germany_Schaefer>
> <http://de.dbpedia.org/resource/Germany http://dbpedia.org/ontology/wikiPageDisambiguates http://de.dbpedia.org/resource/Germany_Schulz>
> <http://de.dbpedia.org/resource/Germany http://dbpedia.org/ontology/wikiPageDisambiguates http://de.dbpedia.org/resource/Germany_Smith>
> <http://de.dbpedia.org/resource/Germany http://dbpedia.org/ontology/wikiPageDisambiguates http://de.dbpedia.org/resource/Jim_Germany>
> <http://de.dbpedia.org/resource/Germany http://dbpedia.org/ontology/wikiPageDisambiguates http://de.dbpedia.org/resource/Reggie_Germany>
> <http://de.dbpedia.org/resource/Germany http://dbpedia.org/ontology/wikiPageDisambiguates http://de.dbpedia.org/resource/Willie_Germany>
> <http://de.dbpedia.org/resource/Germany http://dbpedia.org/ontology/wikiPageDisambiguates http://de.dbpedia.org/resource/Made_in_Germany>

Additionally, http://de.dbpedia.org/resource/Deutschland does not contain any label as "Germany". 

> <http://de.dbpedia.org/resource/Deutschland http://www.w3.org/2004/02/skos/core#altLabel Deutsch>
> <http://de.dbpedia.org/resource/Deutschland http://www.w3.org/2004/02/skos/core#altLabel DE>
> <http://de.dbpedia.org/resource/Deutschland http://www.w3.org/2004/02/skos/core#altLabel Deutschland>
> <http://de.dbpedia.org/resource/Deutschland http://www.w3.org/2004/02/skos/core#altLabel DEU>
> <http://de.dbpedia.org/resource/Deutschland http://www.w3.org/2004/02/skos/core#altLabel BR Deutschland>
> <http://de.dbpedia.org/resource/Deutschland http://www.w3.org/2004/02/skos/core#altLabel Bundesrepublik Deutschland>
> <http://de.dbpedia.org/resource/Deutschland http://www.w3.org/2004/02/skos/core#altLabel Teutschland>
> <http://de.dbpedia.org/resource/Deutschland http://www.w3.org/2004/02/skos/core#altLabel Sprachen in Deutschland>
> <http://de.dbpedia.org/resource/Deutschland http://www.w3.org/2004/02/skos/core#altLabel Federal Republik of Germany>
> <http://de.dbpedia.org/resource/Deutschland http://www.w3.org/2004/02/skos/core#altLabel Vereinigtes Deutschland>
> <http://de.dbpedia.org/resource/Deutschland http://www.w3.org/2004/02/skos/core#altLabel Ed>

Although, our search by context shows dbr:Deutschland as a possible candidate, it fails because there is not any similarity with Germany.

Finally, I thought quite interesting this issue.

RicardoUsbeck commented 6 years ago

Well but you are right, the endpoint and task are mixed up => won't fix