RE: summary of state-of-play

Dear al,

 

a bit late (30minutes to the meeting), but my small two cents on the mail from Philipp (thanks a lot for the useful summarization!).

 

1.       Here I take no strong part. I see advantages in both (I would agree with John, but in general, I tend to over-specify, so the “self-analysis psychologist” in me tell I should appreciate more the advantages of this simplification :-D ). 

2.       Here I agree with John more neatly. I’ve two problems here. I don’t feel confident with the scheme represented in the doc (I’m no more sure of the interpretation of sense and concept this way), and I agree with the fact subsumes may be misleading, suggesting some broader/narrower interpretation (though I agree that the meaning of subsume is a different shade of this, and could actually fit your interpretation).

3.       Here too I agree on the fact that if lexicalConcepts are a specific subclass of skos:Concept, then we may have specific properties for representing our constructs.

 

Regarding point 3, there could be in effect a total inversion of direction, if we take one step back, and we ask ourselves: “what is actually a LexicalConcept?” Has it any feature that really makes it any different from a generic skos:Concept, if not the fact of being used to produce some semantic structure (such as synsets) in a lexical resource? The answer may be, incredibly: “no, there’s no difference”, and thus we may agree to use more standard properties for linking skos:Concepts to anything, but also then we do not have to use this LexicalConcept class. If that is the case, maybe making explicit the use in some lexical resource of skos:concepts to represent any semantic structure (that is, to at least tell that “yes, this lexical resource has a semantic hierarchical structure and not only lexical entries”), should be addressed at metadata level only (there’s a ontolex module foreseen for that, in case). 

Other (non-exclusive) option: we may foresee a specific subclass of skos:Scheme, with some additional features (still, if we feel any as important).

Well I hope here I’m not causing more problems than solutions, but I think answering these questions would speed up modelling, as we would end up with clear modelling exigencies, which only need to be implemented.

 

Last technical comment: we shouldn’t use skos:exactMatch to link skos:Concepts (or any subclass of it) to owl ontology entities, unless we want any reasoner to infer that all these linked entities are also skos:Concepts; exatchMatch is in fact children of skos:semanticRelation, and thus domain and range are set to skos:Concept. Having such kind of inference may not be desired (even solely for the entropic generation of unwanted, even if harmless, triples).

 

Cheers,

 

Armando

 

From: johnmccrae@gmail.com [mailto:johnmccrae@gmail.com] On Behalf Of John McCrae
Sent: Friday, June 7, 2013 4:35 PM
To: Philipp Cimiano
Cc: public-ontolex@w3.org
Subject: Re: summary of state-of-play

 

 

 

On Fri, Jun 7, 2013 at 1:19 PM, Philipp Cimiano <cimiano@cit-ec.uni-bielefeld.de <mailto:cimiano@cit-ec.uni-bielefeld.de> > wrote:

Dear all,

  apologies for the silence in the last two weeks. I have been extremely busy with the organization of ESWC. Now I am getting back to various activities.

I have written a document that tries to summarizes the current agreement that we have about the core of the ontolex model.
Please find this document attached. Feel free to comment on my view of where we stand so far.

I would like to raise three issues that I believe we have not yet considered so far:

1) Why do we not use the same property (i.e. "denotes" both between the Lexical Sense and the Ontology Entity as well as between the Lexical Entry and the Ontology Entity). This would certainly simplify the model and only require one property more when using the lex. sense as mediator between the lexical entry and the ontology entity. Note that the reason is not a technical one as OWL allows left-recursive and right-recursive property chains, that is, we could say that:  sense o denotes \sqsubseteq denotes in OWL.

I can think of several reasons why not (it is ambiguous, confusing and technically harder to work with). I'm not sure I understand why though? You have the same structure and I don't really see how reusing a name from another structure would make that any "simpler."


2) To make clear that linking to lexical concepts from legacy resources (e.g. Synsets of WordNet) is really optional, I propose that we reverse the directionality of the arrow and add a relation "subsumes" from the Lexical Concept to the Lexical Sense. In this sense, a synset subsumes a certain lexical sense. This makes it clear that the "decoration" (to use a term from Armando) of an ontology with synsets or other lexical concepts is not critical to the main path and really "optional".

 There have been a number of namings for properties suggested, we should collect the options and discuss them.

 

For "subsumes", I don't really like it as it suggests some kind of more/less specific relation so would often be misintepreted.


3) Finally, concerning my relation marked with the three question marks in my document. I really wonder if we should aim at relating a lexical concept to an ontological concept or we should simply be agnostic with respect to how this link is made. Some people might want to use owl:equivalentClass or skos:match etc. So why not leaving this simply open?

Well, it is a key goal of the CG to describe how we link lexical resources (including legacy resources) to ontologies, and this includes the synset/concept hierachy of WordNet. We discussed the use of skos:exactMatch to represent this thoroughly in the last two telcos and the consensus was that we introduce a new property for this (Model 1 here http://www.w3.org/community/ontolex/wiki/Teleconference,_2013.10.05,_15-16_pm_CET)

 

Can we find some time this week to discuss things and put into place a basis model for further discussion?

 

Regards,

John


I propose we skip once more this week's telecon and get back to our normal working mode next week. I will write an email in due time.

In the meanwhile, I would be happy to discuss my summary per email.

Best regards,

Philipp.

-- 
Prof. Dr. Philipp Cimiano
Semantic Computing Group
Excellence Cluster - Cognitive Interaction Technology (CITEC)
University of Bielefeld

Phone: +49 521 106 12249 <tel:%2B49%20521%20106%2012249> 
Fax: +49 521 106 12412 <tel:%2B49%20521%20106%2012412> 
Mail: cimiano@cit-ec.uni-bielefeld.de <mailto:cimiano@cit-ec.uni-bielefeld.de> 

Room H-127
Morgenbreede 39
33615 Bielefeld