Re: Linking to semiotics.owl

Dear Philipp, sorry for the late reply, and thanks for summarizing the issues.

On Oct 18, 2013, at 5:35:15 PM , Philipp Cimiano <cimiano@cit-ec.uni-bielefeld.de> wrote:

> Dear all,
> 
>  we have all agreed that we want to have links to other models, that these links should be formal (in the OWL sense) but that the mappings should not be part of the core model, but rather outsourced in a "linking module". The links would thus serve mainly for documentation purposes, but would not play a major role in the actual use of the model.

Sure, however I suggest to simply say that any possible role (major or not) in the actual use of the model is of course mandated to the linking module.

> 
> At least this is the understanding I have from the status of our discussion.
> 
> I will now make a proposal for formal linking to semiotics.owl that I would like to discuss and agree upon (or not).
> 
> In semiotics.owl, an "expression" is defined as: "Any information that either dul:expresses a Meaning or denotes a Reference". To some extent, this definition is quite vague and in my view we can commit totally to the view that a ontolex:LexicalEntry is a kind of semio:Expression". 
> 

Yes, that's what I have been suggesting. Notice that informal definitions in semiotics.owl just provide the explanation of the axioms in the ontology. For the intuition, there are examples provided. I did this in order to avoid taking theoretical positions on what is the "real nature of" things types as expressions, meaning, or references. The core of semiotics is relational, not taxonomic. What counts is what is assumed to denote what, and to express [a conceptualization for] it: "Barack_Obama" may denote http://dbpedia.org/resource/Barack_Obama with the type http://dbpedia.org/ontology/OfficeHolder, or with a gloss http://example.org/myont/WhoIsObama (with an associated value "Barack Obama was born in … is the President of …"). 
Notice that this is very general, and the restriction e.g. used in lemon-ontolex to lexical senses makes lemon-ontolex a special semiotic case.

> If we look at our main path, a LexicalEntry is connected through the relation ontolex:denotes to an Ontology Entity. By this path, a Lexical Entry fullfills exactly the definition of an expression in semiotics.owl, so it seems justified and no over-commitment to link ontolex:LexicalEntry to semio:Expression through a subClassOf relation. We are clearly not making any unjustified commitments here. 
> 

Ok

> As a corollary, it would seem natural to formally link ontolex:denotes to semiotics:denotes through an equivalentProperty statement (or a SubclassOf statement?). The consequence of linking with an equivalentProperty statement would be that ontolex:denotes would then have semiotics:Reference as range. This might seem critical in the light that we never wanted to fix the range of "denotes". 

I think SubPropertyOf is more appropriate. semiotics:Reference includes things that are not necessarily ontology entities, for the abovementioned reason (semiotics.owl does not commit to any particular theory of reference). An EquivalentProperty statement would make us infer that.

> 
> However, our current definition of ontolex:denotes is:
> 
> "The relation between a lexical entry and the logical predicate in an ontology that represents its meaning and has some denotational / model-theoretic semantics"
> 
> And semio:denotes is defines as follows:
> 
> "A relation between expressions and anything (including expressions). 
> It can be used to talk about e.g. entities denoted by proper nouns: the proper noun 'Leonardo da Vinci' denotes the person Leonardo da Vinci; as well as to talk about sets of entities that can be described by a common noun: the common noun 'person' denotes the collection of all persons in a domain of discourse. In OWL2, punning can be used to represent denotation of concept names with owl class extensions, e.g. 'mouse' denotes owl:Class:Mouse."
> 
> It seems to me that both relations semio:denotes and ontolex:denotes are fairly compatible, with ontolex:denotes being more specific requiring a "LexicalEntry" in its domain. From this perspective if makes sense to formally link ontolex:denotes to semio:denotes thorugh a subPropertyOf relation IMHO. Interestingly, our definition is even more specific. While semio:denotes says the range is "anything it can be used to talk about", we are more specific requiring the range to be "a logical predicate (or symbol I would say) in an ontology that represents its meaning and has some model-theoretic semantics".  
> In sum, I feel that linking to semio:denotes is absolutely warranted and actually no overcommitment in any way.

I agree, as I suggest above, lemon-ontolex can be seen as a specialized semiotic model applied to the represention of lexical relations, then a SubPropertyOf linking is justified.

> 
> Now, let's consider the definition of "Reference" in semiotics.owl
> 
> A semio:Reference is "Anything that isDenotedBy an Expression, or that hasInterpretation some Meaning".  If we thus say indirectly that the range of ontolex:denotes is semio:Reference via the subProperty linking to semio:denotes, there is clearly no problem here as the above defnition is rather uncritical. I neither see any overcommitment here.
> 
> So my proposal is that in a separate linking module we add the following links for documentation purposes:
> 
> ontolex:denotes subPropertyOf semio:denotes
> ontolex:LexicalEntry subClassOf semio:Expression
> 

Ah ok, fully agreed

> As such, it would follows that
> 
> ontolex:denotes range semio:Reference
> 
> But as I argue above this is not critical in any way and even fullfills our intuitions.
> 

+1

> 
> I will leave it here for now, we still have to discuss about a link to semio:Meaning and hasConceptualization, but let us first agree on the above and then continue.
> 

For that, please see my forst comment. Also semio:Meaning is fairly general, and does not commit to any particular theory of meaning. Therefore, we can assume that meanings can be any entity intended as the conceptualization of a reference denoted by an expression (or, according to the use case, as expressed by an expression that denotes a reference). 

Such entities include lexical senses, synsets, semantic frames, semantic roles, concepts, ontology classes and relations, but also glosses, paraphrases, definitions, and even cognitive objects. Of course, each case of meaning establishes a special "semiotic game", and lemon-ontolex does just that in considering lexical senses for lexical entries. 

Accordingly, I propose to use SubClassOf and SubPropertyOf statements for the linking.

> I know we sort of agreed on this already, but Guido mentioned in his answer to the poll that we should have a deeper look at the links to make sure that we are not making any commitments that we do not want. Given what I said above, I do not think that we are making any unwarranted commitments.

Indeed
Ciao
Aldo

> 
> I am happy to hear your point of view.
> 
> Best regards,
> 
> Philipp.
> 
> 
> 
> 
> -- 
> 
> Prof. Dr. Philipp Cimiano
> 
> Phone: +49 521 106 12249
> Fax: +49 521 106 12412
> Mail: cimiano@cit-ec.uni-bielefeld.de
> 
> Forschungsbau Intelligente Systeme (FBIIS)
> Raum 2.307
> Universität Bielefeld
> Inspiration 1
> 33619 Bielefeld