Kenny Z. Guan wrote:
> Hi.
>
> I found that the ontology-based data has this information: (with
> "entity" and "type" pair)
>
> Thomas Jefferson and slavery Place
> Thomas Jefferson and slavery Person
> Thomas Jefferson and slavery Country
> Thomas Jefferson and slavery Populated Place
> Thomas Jefferson and slavery Person
> Thomas Jefferson and slavery Person
> Thomas Jefferson and slavery Person
> Thomas Jefferson and slavery Office Holder
>
> can the "Thomas Jefferson and slavery" be a person and a country??
>
> Ken
Things like this happen when the infoboxes are screwed up in a
wikipedia article. In this article, for instance, there are infoboxes
for ""Jefferson the President" (Office Holders) and for the U.S.
Declaration of Independence, two of his slaves and the Louisiana
Purchase. The ontology-based data is driven by infoboxes, so this will
cause the extractor to do the wrong thing.
If you go looking for bad data in dbpedia, you will certainly find
it. However, considering the hundreds of millions of facts about the
millions of entities documented in wikipedia, I'm much more impressed
with the quality of the data than I am with its failings.
That said, scalable reasoning systems based on linked data will need
the ability to manage incorrect facts, encapsulate fiction, and a
number of capabilities that are just barely discussed in the semweb
literature.
At Ontology2 we're starting to develop tools for handling this. For
instance, we've (within limited problem domains) created files that
merge type data from dbpedia and freebase. One logical thing is to
throw out the cases that are obviously crazy, like the one above. I
think everyone would agree that "Person" and "Place" are disjoint
categories, for instance, even though you'll find cases that are not
so simple: for instance, a newspaper (as most people think of it) is
both a "Creative Work" and a "Organization", so you shouldn't assert
that those are disjoint.
On the other hand, there are a lot of "People" in wikipedia w/o
infoboxes, and it's a pretty easy thing to assert the "Person" type on
dbpedia entities that Freebase recognizes as human.
------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Dbpedia-discussion mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/dbpedia-discussion
