Advancing the DBpedia ontology

> From: Peter F. Patel-Schneider [mailto:pfpschneider@gmail.com]
> I agree that there are problems with the mappings.  However, how can the
> mappings be fixed without fixing the ontology?

I could ask you a converse question:
** how can you make an accurate ontology without looking at the data?
And to look at the data, you need mappings (if not to execute then to document what you've examined).

But more constructively:

There is a large number of mapping problems independent of the ontology.
E.g. when a Singer (Person) is mapped to Band (Organisation) due to wrong check of a field "background",
I don’t care how the classes are organized, I already hurt that the direct type is wrong.

Of course, having a good ontology would help!
E.g. https://github.com/dbpedia/mappings-tracker/issues/49:
some guy named Admin made in 2010 two props "occupation" and "personFunction" with nearly identical role & history.
- No documentation of course. 
- occupation has 100-250 uses, personFunction has 20-50 uses.
- Which of the two to use?
- More importantly, which have already been used right, and which are wrong?

I suspect that most uses of occupation are as a DataProp, even though it's declared as an ObjectProp.

DBpedia adopts an Object/DataProp Dichotomy that IMHO does not work well.
See http://vladimiralexiev.github.io/pres/20150209-dbpedia/dbpedia-problems-long.html#sec-3-2

Received on Wednesday, 25 February 2015 04:16:07 UTC