On Thu, Mar 7, 2013 at 2:11 PM, Ahmed Sobhi <[email protected]> wrote:
I ended up using the nt version of dbpedia. It takes a lot of time, the
~20m triples took around 2.5h. I still haven't tried the ttl.
Took me 15 mins (~30K triples/s) or so on a low end desktop machine.
Ubuntu, 8G RAM. 64 bit java.
I assume you have a laptop, maybe running 32 bit java.
(And MS Windows.)
This time, I got no error/warning messages. The POS.dat, OPS.dat and
OSP.dat are all around 760MB.
However, I'm still getting weird results.
A query like
SELECT ?s ?p ?o WHERE {?s ?p ?o.} LIMIT 100
works fine
What about the other queries you were trying?
but
select distinct ?Concept where {[] a ?Concept} LIMIT 100
from dbpedia gives a single Concept <http://www.opengis.net/gml/_Feature>
which is correct given the data.
There are lots of occurrences of rdf:type but they all have object
<http://www.opengis.net/gml/_Feature>
grep '#type' mapping.ttl
is 430731 lines.
but
grep '#type' mapping.ttl | grep -v 'http://www.opengis.net/gml/_Feature'
outputs nothing.
DBpedia has other data loaded.
Andy
Something is still wrong. Any ideas?
On Thu, Mar 7, 2013 at 1:39 PM, Andy Seaborne <[email protected]> wrote:
On 07/03/13 08:40, Ahmed Sobhi wrote:
I was loading the .ttl version of the dump
http://downloads.dbpedia.org/**3.8/en/mappingbased_**
properties_en.ttl.bz2<http://downloads.dbpedia.org/3.8/en/mappingbased_properties_en.ttl.bz2>
.
I kept on getting unicode related warnings so maybe that's the reason.
I got a lot (a LOT!) of warnings about bad URIs including a few normal
form (NFC) warnings but no hard unicode warnings (the load would stop).
Did you issue the three different queries against the same running
server? Or did you run the server three times?
In answer to your questions: yes (as far as I remember but I'll double
check), no, 0.2.6.
I'm repeating but with the .nt version of the dump
http://downloads.dbpedia.org/**3.8/en/mappingbased_**
properties_en.nt.bz2<http://downloads.dbpedia.org/3.8/en/mappingbased_properties_en.nt.bz2>which
doesn't give any warnings and I'll let you know what happens
It is useful to check the data before loading - try "riot --validate".
It is faster than loading (I get it parsing at 220K triples/s).
The ntriples route (not Turtle) does less checking - the URI issues
causing the warnings are still there.
Andy
On Wed, Mar 6, 2013 at 11:19 PM, Andy Seaborne <[email protected]> wrote:
On 06/03/13 19:09, Ahmed Sobhi wrote:
Hi,
I'm a jena newbie. I am trying to load the DBpedia's Ontology Infoxbox
properties
http://downloads.dbpedia.org/****3.8/en/mappingbased_**<http://downloads.dbpedia.org/**3.8/en/mappingbased_**>
properties_en.ttl.bz2<http://**downloads.dbpedia.org/3.8/en/**
mappingbased_properties_en.**ttl.bz2<http://downloads.dbpedia.org/3.8/en/mappingbased_properties_en.ttl.bz2>
,
http://wiki.dbpedia.org/****Downloads38#h227-1<http://wiki.dbpedia.org/**Downloads38#h227-1>
<http://**wiki.dbpedia.org/Downloads38#**h227-1<http://wiki.dbpedia.org/Downloads38#h227-1>>into
a Fuseki server running on
my machine.
For that I did the following
1. Create the tdb using the following command
java tdb.tdbloader --loc=dbpedia mappingbased_properties_en.ttl
2. Ran the server using the following
.\fuseki-server.bat --loc=../dbpedia /dbpedia
...
However, when specifying the subject instead, I get the correct
results.
Any ideas why specifying the predicate or the object doesn't work?
Works for me.
Did the load finish cleanly before you started Fuseki?
Did you move the database after loading?
Which version are you using?
It could be the POS and OSP indexes are broken - how big are the POS.dat
and OSP.dat files?
Andy
Note: The exact query works just fine on DBpedia. I'm therefore
suspecting
Fuseki but I'm unable to pinpoint the reason why it's not working
--
Best Regards,
Ahmed Sobhi
http://about.me/humanzz
