[NLP2RDF] [Dbpedia-discussion] ANN: NLP Interchange Format (NIF) 1.0 Spec, Demo and Reference Implementation

[Sebastian Hellmann]

On 11/29/2011 03:52 PM, Christian Chiarcos wrote:
>>> I thought the PTB Tag for Particle was RP and not PR ?
> You're right, that was a typo. Fixed it. Thank you. (Please note that
> the current version is considered a draft, so there may be further
> minor problems to be fixed. The official release will occur in
> mid-2012.)
> (@Sebastian: penn.owl in the NIF repository needs to be updated, too.)
>
I updated the mercurial repository.  Note that I keep a copy of all the 
OLiA ontologies currently used in the repository. They are included in 
the Java resource path, so they stay stable for each release (this also 
means that currently  the downloadable release 1.2 will not be affected 
and still have the error)
>>>   http://www.netestate.de/nlp/ontology.rdf
> I've also taken a look on your ontology, and your data type properties
> can be rendered as OLiA object properties (e.g., agr by hasNumber and
> hasPerson). As for the object properties, the current OLiA modeling
> would represent these as concepts rather than properties (cf.
> http://purl.org/olia/olia.owl#Head, subconcept of
> http://purl.org/olia/olia-top.owl#SyntacticRole). This was necessary
> because annotation schemes differ with respect to the distribution of
> information between edge labels and phrase labels, and in order to
> abstract from this partially arbitrary differentiation, I wanted to
> have both in the same taxonomy, i.e., as objects of properties. Hence,
> your object properties correspond to concepts in OLiA.
>
> If I'm not mistaken, it is still an unsolved problem how this approach
> can be brought together with NIF. In my approach to model linguistic
> corpora in OWL/RDF, syntactic edges would be represented by
> individuals, linked by hasSource and hasTarget properties to the nodes
> they connect, and assigned an object property with the corresponding
> edge label. For NIF, however, this modeling would probably be too
> verbose.
>
> @Sebastian: What is the current treatment of Stanford dependencies ?
> There should be a similar problem.
The dependencies are modeled as Objectproperty between the strings. I 
treat OLiA as a term repository and converted it to a property hierarchy.
see
http://nlp2rdf.lod2.eu/demo.php?text=President+Obama+on+Monday+will+call+for+a+new+minimum+tax+rate+for+individuals+making+more+than+%241+million+a+year.&action=merge&urirecipe=offset&prefix=http%3A%2F%2Fprefix.given.by%2FtheClient%23&service[]=http%3A%2F%2Fnlp2rdf.lod2.eu%2Fdemo%2FNIFStanfordCore&service[]=

<http://prefix.given.by/theClient#offset_31_35_call>
                                                      stanford:Auxiliary 
<http://prefix.given.by/theClient#offset_26_30_will> ;
                                                      stanford:aux 
<http://prefix.given.by/theClient#offset_26_30_will> ;
                                                      stanford:nsubj 
<http://prefix.given.by/theClient#offset_10_15_Obama> ;
                                                      
stanford:NominalSubject 
<http://prefix.given.by/theClient#offset_10_15_Obama> .


<http://purl.org/olia/stanford.owl#nsubj> rdfs:subPropertyOf
<http://purl.org/olia/stanford.owl#NominalSubject> ;
                 rdf:type owl:ObjectProperty .

This has not yet been included in the spec, because it needs some more 
work and discussion.
Regards,
Sebastian


-- 
Dipl. Inf. Sebastian Hellmann
Department of Computer Science, University of Leipzig
Projects: http://nlp2rdf.org , http://dbpedia.org
Homepage: http://bis.informatik.uni-leipzig.de/SebastianHellmann
Research Group: http://aksw.org