Re: A VoCamp Galway 2008 success story

Hi Michael,

Michael Hausenblas wrote:
> Francois,
> 
> Thanks for your feedback and the question. Though I'm not sure what you 
> technically mean with 'my:links is a named graph' :) 
The system output the links in a named graph. See the following example 
in TRiG:

<my:links>
{
  <http://kmi.open.ac.uk/fusion/dblp#document1632795751_264>
  owl:same_as
<http://kmi.open.ac.uk/fusion/dblp#document1ad8378bff1fe32cd13989741b50fe3eaef0db93> 
.
}

We can then describe it as a void:Linkset as I've described below. This 
allows to attach other information such as the author of the linkset, 
the parameters of the algorithm used to generate it, etc.

> I think the answer 
> is simple: indeed we decided to model datasets and linksets 
> independently from each other. The following example from the (not yet 
> publicly available) voiD guide may illustrate this:
> 
> Let's assume the two well-known linked datasets DBpedia and DBLP:
> 
> :DBpedia void:containsLinks :DBpedia2DBLP  .
> 
> :DBpedia2DBLP rdf:type void:Linkset ;
>               void:target :DBLP .
> 
> So, it is a linking *from* DBpedia *to* DBLP; as RDF is a direct graph, 
> this makes sense quite a lot (the subject 'sits' in DBpedia, the object 
> in DBLP).
> 
> IIRC, we had your option in mind as well [1] but decided to go for the 
> current modeling due to the above reasons. Actually, as I think, the two 
> modelings are equivalent, just with reversed directions:
> 
> YOUR model (if I got it right):
> 
>  :dataset1 <- void:target - :links - void:target - > :dataset2
> 
Right.

> voiD model:
> 
>  :dataset1 - void:containsLinks -> :links - void:target - > :dataset2
> 
There's quite a difference between the two models. In the actual voiD 
model it is supposed that the links are included in one of the datasets. 
  This approach raises again the owl:same_as related problems of stating 
logical equivalence of resources as part of the dataset itself. 
Experience in ontology alignment shows that it's wiser to consider the 
linkset as a separate entity pointing to both (or more) datasets. This 
also allows to have more than one linkset for a given pair of datasets 
(why not, I can trust one linkset more than the other...). Finding links 
from a dataset is done by querying on void:Linkset objects pointing to 
this dataset. Actually, this approach is followed by the oddLinker of 
LinkedMDB, a linkset being recorded as a oddlinker:linkage_run entity.


Cheers,
Francois

> 
> Cheers,
>     Michael
> 
> [1] http://community.linkeddata.org/MediaWiki/index.php?MetaLOD#Discussion
> 
> ----------------------------------------------------------
> Dr. Michael Hausenblas
> DERI - Digital Enterprise Research Institute
> National University of Ireland, Lower Dangan,
> Galway, Ireland
> tel. +353 91 495730
> http://sw-app.org
> ----------------------------------------------------------
> 
> François Scharffe wrote:
>> Hi Michael,
>>
>> thank you for the link ;)
>> It seems that the current voiD spec supposes that the linkset (set of 
>> interlinks) is supposed to be included in one of the two interlinked 
>> datasets. I think it would make sense for the linkset to be considered 
>> independently from the two interlinked datasets, something like this:
>>
>> my:links a void:Linkset ;
>>          void:target <dataset1> ;
>>          void:target <dataset2> ;
>>
>> my:links is a named graph containing the set of owl:same_as links. 
>> This would avoiD to bind a dataset with a set of links.
>>
>> Cheers,
>> François
>>
>> Michael Hausenblas wrote:
>>>
>>> Just to let you know. One of the outcomes of the recent VoCamps [1] 
>>> was that we have agreed on a final layout for voiD (Vocabulary of 
>>> Interlinked Datasets). It is now available at [2] - please note that 
>>> the actual (final) namespace will be 'http://rdfs.org/ns/void#' ... 
>>> can't fix everything within two days, right :)
>>>
>>> A more detailed user guide to follow soon!
>>>
>>> Cheers,
>>>     Michael
>>>
>>> PS: A big thanks to the Neologism (http://neologism.deri.ie/) people 
>>> for creating such an awesome tool and John Breslin for the great 
>>> support re rdfs.org!
>>>
>>> [1] http://vocamp.org/wiki/VoCampGalway2008#Outcomes
>>> [2] http://rdfs.org/ns/neologism/void
>>>
>>> ----------------------------------------------------------
>>> Dr. Michael Hausenblas
>>> DERI - Digital Enterprise Research Institute
>>> National University of Ireland, Lower Dangan,
>>> Galway, Ireland
>>> ----------------------------------------------------------
>>>
>>>
>>>
>>
> 
>