How to extract only some terms of DBpedia?

Question

Dbpedia is a extreme source of triples, but is too big for an ontology parser, if you try read a "offline" version of it (http://wiki.dbpedia.org/Downloads351).

I need only some terms of DBpedia like < http://dbpedia.org/resource/Mad_cow >

In my brainstorming, i think:

Create a Sparql and convert the results inside to a new ontology (Jena ResultSetFormatter.toModel i think...), very similar to this question: http://bit.ly/ao9KYP

Anyone have another suggestion?

Thanks, Celso.

Accepted Answer

5

Why not just grab the bits you want? You don't even need sparql:

$ java jena.rdfcat -out N3 \
     http://dbpedia.org/resource/Mad_cow \
     http://dbpedia.org/resource/Mad_Dog
...
<http://en.wikipedia.org/wiki/Mad_cow>
    foaf:primaryTopic <http://dbpedia.org/resource/Mad_cow> .

<http://dbpedia.org/resource/Mad_Dog_%28disambiguation%29>
    dbpprop:redirect <http://dbpedia.org/resource/Mad_Dog> .

<http://dbpedia.org/resource/Mad_cow>
    rdfs:label "Mad cow"@en ;
    dbpprop:redirect <http://dbpedia.org/resource/Bovine_spongiform_encephalopathy> ;
...

However if there is some common condition the concepts meet DESCRIBE is ideal:

DESCRIBE ?s {
    ... some condition for choosing ?s ...
}

Finally, if you really want a large number of triples, it might be more polite to download a dump in n-triples, then filter for the bits you want.

link

answered 10 Oct '10, 17:38

Comment Bot
3.0k●4●9
accept rate: 41%

edited 10 Oct '10, 17:50

reasoning is possible in these results? to get "superclasses" for example?

(10 Oct '10, 17:52) Celsowm

The first version certainly grabs the sameAs itself, but I'm not sure about the object. In the second you need to check, but could add ?s owl:sameAs ?o and add ?o to the project clause if required.

(10 Oct '10, 17:53) Comment Bot

and CONSTRUCT? (http://n2.talis.com/wiki/SPARQL_intro#CONSTRUCT)

(12 Oct '10, 03:55) Celsowm

CONSTRUCT would work too. It's less clear than DESCRIBE, in that it conflates the condition with the bits you want to extract. (Hrm, I see I read 'superclasses' as 'sameAs' above. Replace the latter with the former in my answer :-)

(12 Oct '10, 07:59) Comment Bot

Answer 2

I have had similar problems with the size of DBpedia. A simple solution is to find the downloads that you are interested and filter out the triples using grep. If I am interested in the article category "Butterflies" and I suspect there are useful triples in the article_categories download I can use the following bash script and get a smaller set of triples that contain the references to the URI of interest.

Hope this helps - Pete

#!/bin/bash
grep "<http://dbpedia.org/resource/Category:Butterflies"  article_categories_en.nt  > dbpedia_nt/categories/butterflies.nt

How to extract only some terms of DBpedia?

Follow this question

Upcoming Events

Related questions