Re: DBpedia 3.2 release, including DBpedia Ontology and RDF links to Freebase

Hello John,

John Goodwin wrote:
> 
>> Regarding you arguments:
>>
>> Too tight restrictions: Which ones specifically are too 
>> tight? If the restrictions cause inconsistencies (which they 
>> are likely to do at the moment), then this is a signal a 
>> problem in the DBpedia data. (Which is one of the purposes of 
>> imposing restrictions.)
> 
> I've noticed that properties like "father" have a domain of "British
> Royal or Monarch" and I wonder if this is too tight. Would you not save
> yourself headaches in the future by relaxing that restriction to Person?
> For example if you want to add in "father" information for US presidents
> will you then have to go back and edit your OWL ontology to include US
> presidents in the domain of "father". 
> 
> Furthermore, I understand disjunctions can be expensive when reasoning
> (not sure if that would be the case in the Dbpedia ontology as it
> doesn't use that much extra OWL). 

You are right. The ontology is automatically created and closely fits 
the data (so at the moment it is indeed too restrictive) and in the 
future this will be done by a community process.

To avoid confusion I believe we have to separate two topics/claims here:

1.) DBpedia should not use domain and range.
2.) Currently, the domain/range restrictions are too restrictive.

Depending on the topic "too tight" can be understood differently. As far 
as I understand, Chris talks about the first topic, while you talk about 
the second. I don't fully agree with Chris in that matter, i.e. I think 
that providing domain and range adds value to DBpedia. I do agree with 
your point of view that some of the domain and range restrictions are 
too restrictive at the moment. The latter can be fixed. We can do this 
either when we have a user interface for the ontology mappings or before 
(manually).

>> I think there are many of those. First of all, they allow 
>> checking consistency in the DBpedia data. Having consistent 
>> data allows to provide nice user interfaces for DBpedia. 
> 
> I'm still not sure how domain and range will help check consistency.
> Don't you need OWL disjoints 

OWL disjoints can (and probably will be) added.

> and other information to find
> inconsistencies, unless of course you check all the inferred types for
> the instances? 

Due to the amount of data any reasoning tasks are challenging, but not 
impossible (maybe a challenge for approximate, incomplete inference 
engines; reasoning with large ABoxes etc.).

Kind regards,

Jens

-- 
Dipl. Inf. Jens Lehmann
Department of Computer Science, University of Leipzig
Homepage: http://www.jens-lehmann.org
GPG Key: http://jens-lehmann.org/jens_lehmann.asc

Received on Tuesday, 18 November 2008 10:07:08 UTC