It is indeed used to link entities and not necessarily only those that don't have an external id. External IDs can be problematic if you want to use them in federated SPARQL queries. In theory you would able to compile the IRI from the external ID and the underlying resulver URL, however there are two problems here.. The resolver URL in wikidata allows only one variable (i.e. $1) and various ontologies require additional processing to compile a URI. the simple example being the identifier that in its literal form used the pattern PREFIX:Identifier, whereas a URI it has PREFIX_identifier. This can mostly be resolved using the BIND operator in the SPARQL query, but since this is an expensive operation, more complex federated queries can time out. Also having to parse an external lID, to get its URI feels a bit patchy. Having a property that directly captures a single mapping to an external namespace can thus be quite helpful.
When I explored a mapping property from Wikidata to external IRI namespaces (including ontologies) there were basically three external mapping properties, being OWL;sameAs, schema:sameAs and Skos.exactMatch. OWL sameas dictated that both items should interchangeable, which is not the case. schema:Sameas and SKOS.exact match were the alternatives. Since SKOS comes with more mapping relations [1], I proposed the property exact match a few years back. Since its acceptance, we have been successfully using it running federated queries to and from Wikidata. Initially, I intended to propose the other mapping relationships from SKOS as similarity properties in Wikidata, since I didn't run into the proper use case yet, but it is still on my radar.
The property P2888 has been quite functional since, if I may say.
[1] https://www.w3.org/TR/skos-reference/#mapping
On Sat, Sep 22, 2018 at 10:09 AM Ettore RIZZA ettorerizza@gmail.com wrote:
@Andra Waagmester: I am a little disconcerted by the property P288 "exact match" https://www.wikidata.org/wiki/Property:P2888. I see it is mostly used to link entities, not properties, and I can't figure out how it differs from an external id (unless it's just a convenient way of linking concepts to databases that do not have an external id in Wikidata?)
On Sat, 22 Sep 2018 at 15:55, Peter F. Patel-Schneider < pfpschneider@gmail.com> wrote:
It is indeed helpful to link the Wikidata ontologies to other ontologies, particularly ones like the DBpedia ontology and the schema.org ontology. There are already quite a few links from the Wikidata ontology to several other ontologies, using the Wikidata equivalent class and property properties. Going through and ensuring that every class and property, for example, in the DBpedia ontology or the schema.org ontology is the target of a correct (!) link would be useful. Then, as you indicate, it is not so hard to query Wikidata using the external ontology or map Wikidata information into information in the other ontology.
The Wikidata ontology is much larger (almost two million classes) and much more fine grained than most (or maybe even all) other general-purpose ontologies. This is appealing as one can be much more precise in Wikidata than in other ontologies. It does make Wikidata harder to use (correctly) because to represent an entity in Wikidata one has to select among many more alternatives.
This selection is harder than it should be. The Wikidata ontology is not well organized. The Wikidata ontology has errors in it. There is not yet a good tool for visualizing or exploring the ontology (although there are some useful tools such as https://tools.wmflabs.org/bambots/WikidataClasses.php and http://tools.wmflabs.org/wikidata-todo/tree.html).
So it is not trivial to set up good mappings from the Wikidata ontology to other ontologies. When setting up equivalences one has to be careful to select the Wikidata class or property that is actually equivalent to the external class or property as opposed to a Wikidata class or property that just happens to have a similar or the same label. One also has to be similarly careful when setting up other relationships between the Wikidata ontology and other ontologies. As well, one has to be careful to select good relationships that have well-defined meanings. (Some SKOS relationships are particuarly suspect.) I suggest using only strict generalization and specialization relationships.
So I think that an effort to completely and correctly map several external general-purpose ontologies into the Wikidata ontology would be something for the Wikidata community to support. Pick a few good external ontologies and put the needed effort into adding any missing mappings and checking the mappings that already exist. Get someone or some group to commit to keeping the mapping up to date. Announce the results and show how they are useful.
Peter F. Patel-Schneider Nuance Communications
On 9/22/18 4:28 AM, Maarten Dammers wrote:
Hi everyone,
Last week I presented Wikidata at the Semantics conference in Vienna ( https://2018.semantics.cc/ ). One question I asked people was: What is
keeping
you from using Wikidata? One of the common responses is that it's quite
hard
to combine Wikidata with the rest of the semantic web. We have our own
private
ontology that's a bit on an island. Most of our triples are in our own
private
format and not available in a more generic, more widely use ontology.
Let's pick an example: Claude Lussan. No clue who he is, but my bot
seems to
have added some links and the item isn't too big. Our URI is http://www.wikidata.org/entity/Q2977729 and this is equivalent of http://viaf.org/viaf/29578396 and http://data.bibliotheken.nl/id/thes/p173983111 . If you look at http://www.wikidata.org/entity/Q2977729.rdf this equivalence is
represented as:
<wdtn:P214 rdf:resource="http://viaf.org/viaf/29578396%22/%3E <wdtn:P1006 rdf:resource="
http://data.bibliotheken.nl/id/thes/p173983111%22/%3E
Also outputting it in a more generic way would probably make using it
easier
than it is right now. Last discussion about this was at https://www.wikidata.org/wiki/Property_talk:P1921 , but no response
since June.
That's one way of linking up, but another way is using equivalent
property (
https://www.wikidata.org/wiki/Property:P1628 ) and equivalent class ( https://www.wikidata.org/wiki/Property:P1709 ). See for example sex or
gender
( https://www.wikidata.org/wiki/Property:P21) how it's mapped to other ontologies. This won't produce easier RDF, but some smart downstream
users
have figured out some SPARQL queries. So linking up our properties and
classes
to other ontologies will make using our data easier. This is a first
step.
Maybe it will be used in the future to generate more RDF, maybe not and
we'll
just document the SPARQL approach properly.
The equivalent property and equivalent class are used, but not that
much. Did
anyone already try a structured approach with reporting? I'm considering parsing popular ontology descriptions and producing reports of what is
linked
to what so it's easy to make missing links, but I don't want to do
double work
here.
What ontologies are important because these are used a lot? Some of the
ones I
came across:
- https://www.w3.org/2009/08/skos-reference/skos.html
- http://xmlns.com/foaf/spec/
- http://schema.org/
- https://creativecommons.org/ns
- http://dbpedia.org/ontology/
- http://vocab.org/open/
Any suggestions?
Maarten
Wikidata mailing list Wikidata@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikidata
Wikidata mailing list Wikidata@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikidata
Wikidata mailing list Wikidata@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikidata