Dbpedia Datasets
Leigh Dodds
I've just posted a version of this to the public-lod mailing list, but
thought I'd include it here too. I've just published some new dbpedia
datasets into Kasabi:
* Dbpedia Core: http://kasabi.com/dataset/dbpedia-core
* Dbpedia Links: http://kasabi.com/dataset/dbpedia-links
* Dbpedia: http://kasabi.com/dataset/dbpedia
The first two are just the dbpedia 3.7 core and external linking
datasets, the last combines or "layers" together those datasets into a
single dataset.
So you can now query dbpedia 3.7 in Kasabi, and have a little more
choice over whether you're just interested in the core dbpedia data or
whether you want to include additional external links. The links are
variable in their quality so there are reasons why you may want to
ignore them.
To help organize datasets that are purely about links, I've also added
a new "Linking" category [1]. You can add this as a category to your
dataset to identify whether its a linkset. We may look at other ways
to handle this in future, but a category works well for now.
I'm also considering other dbpedia subsets that might be more useful
or accessible for people. E.g. dbpedia people and dbpedia places. Some
of these subsets could be normalized to use alternate vocabularies.
Let me know if you have any thoughts or questions.
Cheers,
L.
[1]. http://kasabi.com/browse/datasets/results/og_category%3A5603
--
Leigh Dodds
Product Lead, Kasabi
Mobile: 07850 928381
http://kasabi.com
http://talis.com
Talis Systems Ltd
43 Temple Row
Birmingham
B2 5LS
Bob Ferris
> Some
> of these subsets could be normalized to use alternate vocabularies.
+1
Cheers,
Bo
Alfredas Chmieliauskas
But a query:
select distinct ?rel where {
?c1 a <http://dbpedia.org/ontology/Company> .
?c2 a <http://dbpedia.org/ontology/Company> .
?c1 ?rel ?c2 .
} limit 100
breaks the thing (get not well formed xml or an empty json). The query
works on live.dbpedia.org
Alfredas
Leigh Dodds
Apologies for the slow response
On 28 October 2011 13:11, Alfredas Chmieliauskas <alfr...@gmail.com> wrote:
> Great news.
>
> But a query:
> select distinct ?rel where {
> ?c1 a <http://dbpedia.org/ontology/Company> .
> ?c2 a <http://dbpedia.org/ontology/Company> .
> ?c1 ?rel ?c2 .
> } limit 100
>
> breaks the thing (get not well formed xml or an empty json). The query
> works on live.dbpedia.org
Looks like its hitting a time-out issue on our servers. There's a
release going out today which should help reduce occurence of
truncated responses. We're also investigating whether we can move the
dbpedia datasets to different hosting setup to improve performance.
Cheers,
L.
Alfredas Chmieliauskas
I can imagine that this query could be resource-intensive. I think one
thing dbpedia do is caching the query results. Another thing of course
is the error reporting. It would be great to get an error msg instead
of truncated response that does not parse.
And a third thing: it would be great to have a html or text format in
the test query page. For now there are only json and csv which require
external viewers to check the query output.
Otherwise - great stuff!
Alfredas