Re: current draft of the spec

Hi,

I am a bit confused about what we are proposing here. It has been discussed
many times that a word (lexical entry) has a number of word (lexical)
senses and that these are defined by reference to an ontology. As such each
lexical sense is the pair (lexical entry, ontology entity), i.e., each
sense is unique to its entry. As such I am puzzled as to why we are
discussing this axiom:

\forall x,y,z,s LexicalVariant(x) \wedge variantSource(x,y) \wedge
targetSource(x,z) \wedge sense(y,s) \rightarrow sense(z,s)

This would lead to contradictions as the property sense should be inverse
functional.

As best as I understand, the goal is to differentiate between "variants
that have the same sense(s) and reference(s)" and "variants that are
extensionally equivalent". This is precisely why we introduced the *Lexical
Concept * to differentiate between this level.

Can I suggest the following? We should have following 4 classes of variants
(introducing one new one), as follows:

   - Form variants
   - Lexical variants
   - Terminological (sense) variants
   - Semantic (conceptual) variants

Each of which correspond to the four core classes with the following
axiomatization (where y is {form, lexical, terminological, semantic} and X
is {Form, LexicalEntry, LexicalSense, LexicalConcept})

There is a property yVariant which range X and domain X
There is a class YVariant
∀ x,y,z : YVariant(x) ∧ variantSource(x,y) ∧ variantTarget(x,z) ⇒
yVariant(y, z)

(although I don't think this is easy to capture in OWL)

As such, it is clear that we have the following example of each of these:

   - Form variant: cat/cats
   - Lexical variant: FAO/Food and Agriculture Organization
   - Terminological variant: tuberculosis/phthisis
   - Semantic variant: risotto/rice dish


Regards,
John





On Thu, Jan 30, 2014 at 9:09 PM, Aldo Gangemi <aldo.gangemi@cnr.it> wrote:

> Just a couple of comments, one on using OWL for Philipp’s axioms, see below
>
> On Jan 30, 2014, at 8:19:31 PM , Philipp Cimiano <
> cimiano@cit-ec.uni-bielefeld.de> wrote:
>
>  Hi Elena,
>
>  see below as well...
>
> Am 30.01.14 16:05, schrieb Elena Montiel Ponsoda:
>
> Dear Philipp,
>
> Thanks for this.
> Some comments between lines.
> Best,
> Elena, Jorge, Lupe
>
> El 30/01/2014 9:04, Philipp Cimiano escribió:
>
> Dear all,
>
>  I have been working on the final spec this morning. Please have a look at
> the modified examples.
>
> I also added an example of a Spanish lexicon so that we show how we would
> get interoperability between lexica in different languages, an important
> aspect to hint at briefly I think.
>
> Spanish people: could you please check ;-)
>
> Other than that, I have been trying to define a bit better the types of
> variants that we are considering (as already discussed with Lupe and Jorge
> during our last telco).
>
> I think it would be important to clarify what *we* mean with these things.
> Let me make a proposal for lexical variant and terminological variant. From
> there we can move to semantic variant and translation.
>
> 1) Lexical Variant:
>
> Lexical variants were defined as those variants that are semantically
> coincident (same meaning) but formally different, and which are mainly
> motivated by grammatical requirements, style, and linguistic economy
> (helping to avoid excessive denominative repetition and improving textual
> coherence). With respect to the ontology-lexicon model, two lexical
> variants are different lexical entries that have the same sense(s) and
> reference(s) and are thus semantically equivalent. LexicalVarient thus
> represents a relation between two Lexical Entries.
>
> From our point of view, and in fact you have an example of this in the
> core model specification (see color vs. colour), two lexical variants are
> different forms of the same lexical entry.
>
> Well, as I said LexicalVariant is supposed to be a relation between
> lexical entries and color and colour are two different forms of the same
> LexicalEntry. So I would not say that color and colour are lexical
> variants, would you? In my view they are two forms of the same lexical
> entry.
>
>
> What do you understand by "semantically coincident (same meaning)"? That
> they have the same ontology reference??
> In the case of Terminological variants, would you state the same? Would
> they also be "semantically coincident but formally and also *pragmatically
> *different"?? (See also the definition that we propose for Terminological
> variants below). If we remember correctly, the problem with the previous
> definitions was that we had three levels:
>
>    - formally (different forms)
>     - semantically (different senses)
>     - conceptually (different ontology references)
>
>
>  Well, that is exactly what I am trying to understand, the definition
> "semantically coincident (same meaning)" came from UPM. I am just trying to
> find out what it means and proposing some more precise definitions
>
>
> and a fourth: “pragmatically” (different contexts of usage: jargon,
> technical terms, social connotations, geographically-bound, etc.)
>
>
>  So, we would thus have:
> \forall x,y  LexicalVariant(x) \wedge variantSource(x,y) \rightarrow
> LexicalEntry(y) (expressible in OWL?)
>
>
> yes, if variantSource is used typically for this purpose:
>
> variantSource rdfs:domain LexicalVariant
> variantSource rdfs:range LexicalEntry
>
> otherwise:
>
> LexicalVariant rdfs:subClassOf (variantSource only LexicalEntry)
>
> \forall x,y LexicalVariant(x) \wedge variantTarget(x,y) \rightarrow
> LexicalEntry(y) (expressible in OWL?)
>
>
> similarly:
>
> variantTarget rdfs:domain LexicalVariant
> variantTarget rdfs:range LexicalEntry
>
> otherwise:
>
> LexicalVariant rdfs:subClassOf (variantTarget only LexicalEntry)
>
> Further:
> \forall x,y,z,s LexicalVariant(x) \wedge variantSource(x,y) \wedge
> targetSource(x,z) \wedge sense(y,s) \rightarrow sense(z,s)
>
>
> This is harder, but this property chain should work:
>
> [inverse(targetSource) *o* variantSource *o* sense] subPropertyOf sense
>
> \forall x,y,z,s LexicalVariant(x) \wedge variantSource(x,y) \wedge
> targetSource(x,z) \wedge sense(z,s) \rightarrow sense(y,s)
>
>
> [inverse(variantSource) *o* targetSource *o* sense] subPropertyOf sense
>
>
> The fact that they have the same concept follows from the functionality of
> "reference", i.e.
> \forall s,r1,r2 reference(s,r1) \wedge reference(s,r2) \rightarrow r1=r2
>
>
> and this needs that “reference" be a functional property:
>
> Functional(reference)
>
> We can follow the same OWL patterns for the terminological variant axioms
> :)
> Ciao
> Aldo
>
>  Could you also explain this in words?? ;)
>
>
> Well, the axioms simply say that for the case of a Lexical Variant both
> lexical entries that stand in relation to each other share the same set of
> sense (and consequently the same reference because reference is
> functional). Is this what you understand by lexicalVariant?
>
> Basically, lexical variants would thus be intensionally, semantically and
> pragmatically equivalent.
>
> So I simply give the question back: what does: "semantically coincident
> (same meaning)" mean? I was just making a proposal that we can discard.
>
>
> Do we agree on this understanding of lexical variant?
>
>
> 2) Terminological Variant:
>
> Terminological Variations are relations between LexicalEntries that have
> two (different) senses that however have the same concept as reference. One
> could thus say that the meanings of these lexical entries are extensionally
> equivalent, but differ intensionally and pragmatically in that the lexical
> entries are used in different contexts, domains, have a different register
> or have different pragmatic connotations.
>
> Here we would suggest following a similar structure as the one followed in
> the definition of LexicalVariant to be coherent. In that sense, we would
> propose:
>
> Terminological variants have the same concept as reference, but differ
> formally and pragmatically in that the lexical entries are used in
> different contexts, domains, have a different register or have different
> pragmatic connotations.
>
> With respect to the ontology-lexicon model, a TerminologialVariant
> connects two different lexical senses of two different lexical entries that
> have the same or equivalent ontology references.
>
>
> Fine, so they have different senses, but the same reference.
>
> A question: is TerminologicalVariant a relation between senses (with the
> same reference) or lexical entries (having different senses with the same
> reference)???
>
>
>
> So we have again:
>
> \forall x,y  TerminologicalVariant(x) \wedge variantSource(x,y)
> \rightarrow LexicalEntry(y) (expressible in OWL?)
> \forall x,y TerminologicalVariant(x) \wedge variantTarget(x,y) \rightarrow
> LexicalEntry(y) (expressible in OWL?)
>
> Further:
>
> \forall x,y,z \exists LexicalVariant(x) \wedge variantSource(x,y) \wedge
> targetSource(x,z) \wedge sense(y,s1) \rightarrow \exists s2,r sense(x,s2)
> \wedge s1 != s2 \wedge reference(s1,r) \wedge reference(s2,r)
> And the converse axiom:
>
> \forall x,y,z \exists LexicalVariant(x) \wedge variantSource(x,y) \wedge
> targetSource(x,z) \wedge sense(x,s1) \rightarrow \exists s2,r sense(y,s2)
> \wedge s1 != s2 \wedge reference(s1,r) \wedge reference(s2,r)
>
> Do we agree on this understanding of terminological variant?
>
> Enough ontolex for me today ;-)
>
> Looking forward to your comments.
>
> 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
>
>
>
> --
> Elena Montiel-Ponsoda
> Ontology Engineering Group (OEG)
> Departamento de Inteligencia Artificial
> Facultad de Informática
> Campus de Montegancedo s/n
> Boadilla del Monte-28660 Madrid, Españawww.oeg-upm.net
> Tel. (+34) 91 336 36 70
> Fax  (+34) 91 352 48 19
>
>
>
> --
>
> 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
>
>
>