On Sun, 1 Apr 2012, Markus Krötzsch wrote:
A very interesting discussion. Some general answers to
* Wikidata does, of course, not intend to implement complex reasoning (or any
other algorithm that qualifies as "complex").
* If useful for serving its requirements, Wikidata will not exclude modelling
features just because they are also supported in OWL ;-) For example, it
could be useful to say that Wikidata item describes the same as an external
resource, which can be done in OWL using sameAs. Many communities could use
this for integrating Wikidata information with other Web databases.
* The "reasoning" support in Wikidata will not in general limit the modelling
support in Wikidata: it might be possible to say something that has a formal
meaning in OWL, even if this formal meaning is not relevant for query
answering in Wikidata (sameAs with external resources is a possible example,
since Wikidata would surely not pull data from these sources for internal
* Wikidata will support various export formats, which have more or less
native support for certain modelling features. We will use whatever
expressivity is available in the given format to describe the Wikidata
information as accurately as possible. This might again lead to some OWL
constructs being used in RDF/OWL exports. All Wikidata content will have a
formal meaning, and we will draw from existing experience and standards for
defining this so that it is as widely compatible as possible.
In summary, it is not about endorsing or rejecting a particular ontology
language. We will be open and inclusive with what we support, and user
requirements will be the main guideline for defining "what can be said" in
This sounds good to me. (Not in the least because this would allow data
representations that are optimized for certain classes of knowledge, e.g.,
Topic Maps, or mathematical/physical relationships.)
But it triggers the obvious question: when and how will such discussions
(and decision making) be done in the course of the coming year?
On 01/04/12 08:54, Ivan Herman wrote:
> On Mar 31, 2012, at 11:17 , Jakob Voss wrote:
>> JFC Morfin wrote:
>>> 2. Since we have a W3C expert: what is the best document/book to get
>>> a comprehensive and clear (not too massive) documentation on the
>>> semantic web?
>> You surely don't want to know all about semantic web - especially the
>> Ontology stuff with OWL dialects and entailment regimes is far too
>> academic and won't be part of wikidata because of computational
>> complexity anyway. In short, you should be *very sceptical* and
>> cautious every time you stumple upon anything that requires inference
>> rules. Even trivial inference rules such as those based on owl:sameAs
>> and rdf:type can be problematic in practice! The less inference you
>> assume, the better.
> Let us avoid the all-to-simplistic view that says Semantic Web == OWL:-)
> Indeed, bringing in (OWL) inferencing into the core WD project would be a
> mistake. From the SW stack, RDF, RDFS, and, on a different note, SPARQL and
> maybe RDB2RDF should be the technologies having a role in the project, as
> well and Linked Data patterns in general.
> That being said, it is probably good to have the vocabularies being used in
> WD be properly defined/described. If *somebody else* wants to do
> inferencing, for example, we should not stand in the way.
>> I can recommend the "Linked Data Patterns" book by Dodds and Davis:
> Indeed. That is a great one, too