Hello Peter,

thank you very much for your inputs. The state of the DBpedia ontology is 
certainly an issue.
You can register at [1], ask for editing rights, and go on and make your 
changes. I'd also feel not quite well performing major changes or removing 
classes without some discussion, since it is the effort of others and it is not 
always clear, if somebody actually uses it. 

Maybe, we could organize an ontology enhancement and guidelines workshop at the 
next DBpedia Community Meeting in Leipzig [2].

[1] http://mappings.dbpedia.org/index.php?title=Special:UserLogin
[2] http://wiki.dbpedia.org/meetings/Leipzig2014


On Apr 10, 2014, at 10:03:05 PM , Patel-Schneider, Peter 
<peter.patel-schnei...@nuance.com> wrote:

> 
> On Apr 10, 2014, at 6:14 AM, Marco Fossati <hell.j....@gmail.com> wrote:
> 
>> Hi Peter,
>> 
>> Thank you for your detailed report.
>> 
>> The DBpedia ontology is (a) crowdsourced and (b) follows a data-driven 
>> approach. Classes and properties are mainly derived from the actual data 
>> coming from different Wikipedia chapters.
> 
> I don't think that this last is true.  For example, there are about 250 
> classes in the ontology that do not have any instances, at least in the data 
> that I have examined, and more that have no instances that are not instances 
> of any of their subclasses.   There are also a number of places where the 
> ontology organization does not match the information in Wikipedia. 
> 


SELECT DISTINCT ?type WHERE {?type a owl:Class. FILTER NOT EXISTS {?subject 
a/rdfs:subClassOf* ?type.} }

As for the English DBpedia dataset there are 142 unused classes:

http://dbpedia.org/sparql?default-graph-uri=http%3A%2F%2Fdbpedia.org&query=SELECT+DISTINCT+%3Ftype+WHERE+%7B%3Ftype+a+owl%3AClass.+FILTER+NOT+EXISTS+%7B%3Fsubject+a%2Frdfs%3AsubClassOf*+%3Ftype.%7D+%7D&format=text%2Fhtml&timeout=30000&debug=on

As Marco said, you'd also need to consider other Language Chapters that use the 
same ontology. But obviously there are some classes needless, redundant, badly 
described, or just wrong.

>> Those are the main reasons of the issues you mentioned.
> 
> The DBpedia ontology looks much more like a unreviewed crowdsourced artifact 
> than an artifact that matches either the information in Wikipedia or in 
> DBpedia.
>> 
>> It would be great if you could contribute a deep analysis and detect the 
>> inconsistencies.
> 
> Well, there are no formal inconsistencies in the DBpedia ontology, as it is 
> too inexpressive to have inconsistencies.  All that can be done is pointing 
> out where the ontology does not appear to match either the normal definitions 
> of the categories or the definitions found by examining Wikipedia information 
> and differences between different parts of the ontology.  I do have a list of 
> all the empty classes, but this information is available elsewhere.  
> 
> The analysis that I sent out yesterday lists quite a number of deficiencies 
> in the ontology.   This analysis should be good enough to serve as a start on 
> fixing the ontology.
> 
>> In this way, we could clean the ontology up and provide rock solid 
>> semantics.
> 
> I agree that this would be a good idea.  However, this should be an iterative 
> process, and should not depend on a complete analysis.
> 
>> As you already mention lots of examples, a brand new ontology and exact 
>> deltas with the current one would be highly beneficial.
> 
> Well, producing a new ontology is work, and it would be nice that this work 
> has an effect.  This is why I was wondering who else was interested in 
> improving the ontology.
> 
>> 
>> Cheers!
> 
> 
> peter
> 

Best regards
Magnus

-- 
Magnus Knuth

Hasso-Plattner-Institut für Softwaresystemtechnik GmbH
Prof.-Dr.-Helmert-Str. 2-3
14482 Potsdam

Amtsgericht Potsdam, HRB 12184
Geschäftsführung: Prof. Dr. Christoph Meinel

tel:     +49 331 5509 547
email:   magnus.kn...@hpi.uni-potsdam.de
web:     http://www.hpi.uni-potsdam.de/
webID:   http://magnus.13mm.de/


------------------------------------------------------------------------------
Put Bad Developers to Shame
Dominate Development with Jenkins Continuous Integration
Continuously Automate Build, Test & Deployment 
Start a new project now. Try Jenkins in the cloud.
http://p.sf.net/sfu/13600_Cloudbees
_______________________________________________
Dbpedia-discussion mailing list
Dbpedia-discussion@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dbpedia-discussion

Reply via email to