Re: Multi-referenced blank nodes and Thea - was: OWL restrictions lost in Turtle export

Hi Jan,

FYI, I've managed to get the Turtle roundtrip running. Exporting and
importing works for the fast majority of models (>> 80% with about 700
models).
Roundtripping with Turtle seems to work fine.

The model I've sent is one of the exceptions. The check for a
particular OWL restriction (using Thea) fails when the
multi-referenced blank nodes have the same node-ID, but succeeds when
I manually create different blank nodes with different IDs. I have yet
to figure out what the exact problem is.

Cheers,
Jochem



On Fri, Jun 28, 2013 at 6:28 PM, Jan Wielemaker <[email protected]> wrote:
> On 06/28/2013 05:18 PM, Jochem Liem wrote:
>>
>> Hi Jan,
>>
>> You are absolutely right about the blank nodes. I completely read over
>> them.
>>
>> I'm still having a problem when loading (or saving) the Turtle files,
>> which is related to the Turtle output reusing equivalent blank nodes.
>>
>> I cannot load the model I've previously sent normally (due to the
>> Restriction not appearing in the OWL abstract syntax). However, when I
>> create differently named copies of each of the multi-referenced
>> blank-nodes this problem disappears. I think this might have to do
>> with the Thea library consuming triples when converting them to OWL
>> abstract syntax.
>>
>> I think I either have to make sure that equivalent blank nodes are not
>> reused when a file is saved, or make sure that multi-referenced blank
>> nodes result in separate blank nodes when a Turtle file is loaded.
>> I've been trying the rdf_load/2 option blank_nodes(noshare), but this
>> does not seem to solve the problem.
>
>
> You mean using rdf_load/2 using blank_nodes(noshare) on the original
> RDF/XML?  If I recall well, there was a nodeID thingy in the original.
> That suggests sharing as well.
>
>
>> Do you have any advice about how I should proceed?
>
>
> I guess find out how the triple set between the loaded XML/RDF version
> differs from the Turtle loaded one.  It shouldn't.  Could it be that
> Thea is relying or triple ordering?  Such assumptions would be wrong.
>
>         Cheers --- Jan
>
>
>>
>> Cheers,
>> Jochem
>>
>>
>> --
>> =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
>> Jochem Liem, MSc.
>> Informatics Institute
>> Faculty of Science
>> University of Amsterdam
>> http://www.science.uva.nl/~jliem/
>>
>> Mobile: +31 (0)6  4321 9992
>>
>> Visitor address:
>> Science Park 904, C2.248
>> 1098 XH Amsterdam
>>
>> Mailing address:
>> Postbus 94323
>> 1090 GH Amsterdam
>> The Netherlands
>> =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
>>
>



-- 
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Jochem Liem, MSc.
Informatics Institute
Faculty of Science
University of Amsterdam
http://www.science.uva.nl/~jliem/

Mobile: +31 (0)6  4321 9992

Visitor address:
Science Park 904, C2.248
1098 XH Amsterdam

Mailing address:
Postbus 94323
1090 GH Amsterdam
The Netherlands
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=