In Wikidata, the subclass hierarchy and the way that properties are used is unmanaged and contradictory. Furthermore, Wikidata added statement qualifiers which put the meaning of any statement in doubt. For example, there is a property "use" https://www.wikidata.org/wiki/Property:P366 . If someone qualifies a statement with "use X" what does that mean? Is the statement no longer generally true? Should it be omitted? The semantics of Wikidata qualifiers have not been defined and and won't be enforced. It's left up to users to invent their own meanings. (In this way, Wikidata is still a lot like the prose in Wikipedia.)
We need more "curated" projects like DBpedia to do the work of maintaining a coherent subclass hierarchy and to take a conservative approach to statements with qualifiers (omitting most such statements unless the qualifier is unambiguous).
- Jeff
On 2018/05/05 15:02, David Abián wrote:
Since the subject has come out, I leave some general impressions, which aren't necessarily applicable to the people in charge of generating the LOD cloud.
Many DBpedia-centered researchers are truly reluctant to mention Wikidata. Some of them don't want people to know that Wikidata exists, so they continue introducing DBpedia in their talks and papers as the largest knowledge base that is available out there — which is, indeed, no longer true. This isn't hate but an attempt to survive, an attempt to ignore change, to continue working on the same lines of research and "enjoying" the corresponding, sometimes poor, funding.
It's not a matter of triples. The very ideas of both projects are different, and this point is what makes DBpedia potentially obsolete. DBpedia is a non-collaborative project — as we understand collaboration in the Wikimedia movement — that emerged from academia with the aim of *extracting* information from Wikipedia. Similarly to Wikipedia, it can be confusing to talk about DBpedia in the singular because there are several DBpedias, each one mainly oriented, and limited, to a language, and not very well interlinked. There's, however, a single multilingual Wikidata that makes the idea of extracting information from Wikipedia less meaningful. Most relevant structured data are already centralized here, in Wikidata, which *provides* them to Wikipedia. Moreover, the data in Wikidata are referenced... sometimes :), and they are more fine-grained and better structured than those in DBpedia.
Researchers should have nothing to fear from Wikidata, and some of them, mainly the young ones, do start to work on our project. In my humble opinion, we need the help of universities and research centers to fill some gaps and to produce and apply theory. I think these needs should be better communicated to researchers and fears should be mitigated. Our project isn't "that new" today.
Hopefully, Wikidata will appear soon in the LOD cloud... O:)
El 04/05/18 a las 18:33, Maarten Dammers escribió:
It almost feels like someone doesn’t want Wikidata in there? Maybe that website is maintained by DBpedia fans? Just thinking out loud here because DBpedia is very popular in the academic world and Wikidata a huge threat for that popularity.
Maarten
Op 4 mei 2018 om 17:20 heeft Denny Vrandečić <vrandecic@gmail.com mailto:vrandecic@gmail.com> het volgende geschreven:
I'm pretty sure that Wikidata is doing better than 90% of the current bubbles in the diagram.
If they wanted to have Wikidata in the diagram it would have been there before it was too small to read it. :)