Re: UMBEL Publicly Released

Hi Chris,

> Very nice work and great to see that UMBEL is available as Linked Data 
> and already interlinked with various ontologies and datasets.
>
Thanks! Well, at least 2 datasets: Yago (and at the same time, dbpedia) 
and the John Peel Session; and about 10 external ontologies.

But much more is to come; it is just the beginning, particularly since 
it is the first public release.

>> A further 1.5 million named entities have been extracted from 
>> Wikipedia and mapped to the UMBEL reference structure with 
>> cross-links to YAGO [3] and DBpedia [4].
>
> In order to allow people to browse from DBpedia into UMBEL and in 
> order to give Semantic Web crawlers more starting points to crawl 
> UMBEL and its interlinked datasets, it would be great to set RDF links 
> from DBpedia into UMBEL.
Certainly :) Well, the first anchor point between all this stuff are 
Wikipedia URIs. So I think the goal would be to check which Yago URI 
exists into DBPedia and to create a first linkage that way.

Since each Yago resource are representations of the same *thing* as the 
representations of DBpedia, I don't think this would be a problem.
> Would it be possible that you send us the 1.5 million RDF links from 
> UMBEL to DBpedia that you already got so that we can serve them 
> together with DBpedia via our Linked Data interface and SPARQL endpoint?
I will have to produce what I have been talking about above. So I will 
check when I can find some time for this (I am away at the beginning of 
the next week; so this could  be at the end of next week, or tomorrow if 
we are lucky ;) )

> Reading the UMBEL documentation, I noticed that you use the 
> umbel:isLike property to link to DBpedia and the owl:sameAs property 
> to link to YAGO. For instance:
>
> ne:Pfizer umbel:isLike dbpedia:Pfizer
> ne:Pfizer owl:sameAs yago:Pfizer
>
> The UMBEL spec defines umbel:isLike  as
>
> "Additionally, the property umbel:isLike can be used to state that two 
> named entities "likely" have the same identity."
>
> Why do you think that your RDF links to YAGO are more likely to be 
> correct than your links into DBpedia?
You are talking about the schema, right? Well no offense, but it was 
probably to show that two properties (sameAs and isLike) could be use to 
do the linkage. However I do agree that both Yago and DBPEdia should be 
sameAs *or* isLike (since they are based on the same data source, they 
are different representations of the same "thing"). So let me fix this 
schema to resolve this potential ambiguity. (I have done this schema 
months ago, so it probably slipped away from the review :) )
> As all three datasets are derived from Wikipedia and Wikipedia page 
> identifiers should be present in all three datasets, I think we could 
> use owl:sameAs between all three.
I agree.

If you check the named entities resources, they are all sameAs; so this 
is really a matter of schema more than anything else.

Check back later this morning (EST) and this will be resolved.


> Keep on the great work and let's hope that UMBEL develops into an 
> important interlinking hub for the Web of Data.

Thanks!

And for this to happen, we will need feedbacks, suggestions, comments 
and bug reports from everybody.


Take care,


Fred

Received on Thursday, 17 July 2008 13:39:05 UTC