Listing common misevaluations

9 views
Skip to first unread message

ploo...@gmail.com

unread,
Nov 16, 2012, 3:11:01 PM11/16/12
to dbpedia-da...@googlegroups.com
Hi again

I think I may have a suggestion you guys could have already come up with.

While you review the users' evaluations I'm pretty sure there can be a number of patterns emerging where users mark some triples as errorneous but in fact these are not.

For example I have a number of times come accross triple like

dbpedia:something dbprop:yearFounded "2000"^^xsd:integer

Instinctively it seems not right since there is the xsd:Year datatype. However I wasn't really sure whether it actually is an error or maybe this is the intended way it should have been extracted. Thus a set of examples of what people can encouter and some guidelines how to react in common cases could be good and also spare some of your time.

Regards,
Tom

Amrapali J Zaveri

unread,
Nov 17, 2012, 5:30:13 AM11/17/12
to ploo...@gmail.com, dbpedia-da...@googlegroups.com
Hi Tom,

Thanks for your suggestions.

On Fri, Nov 16, 2012 at 9:11 PM, <ploo...@gmail.com> wrote:
Hi again

I think I may have a suggestion you guys could have already come up with.

While you review the users' evaluations I'm pretty sure there can be a number of patterns emerging where users mark some triples as errorneous but in fact these are not.

Yes, we do intend to check the evaluations to ensure they are in fact errors and also to find out the root cause problem. 

For example I have a number of times come accross triple like

dbpedia:something dbprop:yearFounded "2000"^^xsd:integer

Instinctively it seems not right since there is the xsd:Year datatype. However I wasn't really sure whether it actually is an error or maybe this is the intended way it should have been extracted.

It does seems like a datatype error where it should have been year. In general, we have noticed that the most errors lie in the values associated to the dbpprop property. 
 
Thus a set of examples of what people can encouter and some guidelines how to react in common cases could be good and also spare some of your time.

The reason we are doing this campaign is to find such common errors. Although we have found a number of them (as listed in the tree of errors), we hope that others can find other types of errors too. In fact each error type has an associated example and a description so as to help people correctly evaluate similar problems.

Hope that helps !

Thanks.

Regards,
Tom

--
You received this message because you are subscribed to the Google Groups "DBpedia Data Quality" group.
To post to this group, send email to dbpedia-da...@googlegroups.com.
To unsubscribe from this group, send email to dbpedia-data-qua...@googlegroups.com.



Regards,
Amrapali Zaveri Gokhale
Reply all
Reply to author
Forward
0 new messages