Re: Next skype call on 24th of October, 14:00 CET

Hi,

On Tue, Oct 24, 2017 at 7:59 AM, Jorge Gracia <jgracia@fi.upm.es> wrote:

> Hi John, all,
>
> Thanks for your comments. Let me share my view on this, for later
> discussion in today's telco
>
> > Thus, in my interpretation, senses in traditional dictionaries should
> always be modeled as lexical concepts.
>
> I think that there are still good reasons to model "dictionary senses" as
> "lexical senses" in some cases (as people have been done in previous lemon,
> btw). In particular, according to the ontolex specification, translations
> (and other types of sense relations) are established between lexical senses
> but not between lexical concepts. Therefore, bilingual/multilingual
> dictionaries need the creation of lexical senses if we want to represent
> translations explicitly. In such cases, creating a "lexical concept"
> counterpart for the needed lexical senses might result in overloading the
> modelling unnecessarily.
>
Yes, actually if you follow my chart, in the case of a bilingual
dictionary[1] lexical senses would be the correct answer.

Is there are definition? No (or at least not consistently so, as many
entries in bilingual dictionaries have no definitions)
Does this object have relations? Yes
Are there subsumption (is-a) relations? No
Do these relations apply to synonyms? Never (as in a biliingual dictionary,
synonyms have different translations)*
Do these relations apply to homographs? No => LexicalSense

This changes when you have proper definitions (such as how Seppälä defines
a 'definition' [2] as having at least a *genus *and a *differentia*). In
this case there is clearly a 'mental concept' being evoked by the
definition and so a lexical concept must be introduced. For the case of
Petit Larousse, we have in my mind clearly stepped over this boundary and
are talking about these 'mental concepts'.

> I think we need to make a clear and unambiguous definition of lexical
> sense, that precludes its usage without an ontological
> > reference, as this is how it is defined both in the final specification
> and in the OWL code (reference exactly 1).
>
> Although I understand this point, relaxing this restriction, as we did in
> the RDF generation of the Apertium family of bilingual dictionaries, proved
> to be very useful: in short, to represent translations we created
> "artificial" lexical senses (associated to their corresponding lexical
> entries in different languages, and defining translations between them) to
> support the fact that translations were established not between words but
> between word meanings, leaving such lexical senses not linked to external
> references initially. Then, as a later step, we were able to connect some
> of these "orphan" lexical senses to BabelSynsets, while other senses remain
> unconnected. This might be not 100% compliant with the ontolex definition
> but is a natural pattern that emerges when you model "botton-up" from
> lexicon to ontology, and gives you the flexibility of connecting senses
> among them (e.g., via translations) or to external references when
> available (e.g., BabelNet) even if the reference is not known beforehand.
>
> In fact, the idea of "precluding the usage of lexical sense without an
> ontological reference" corresponds to a purely top-down vision, that is,
> lexical senses can be used only when you "know" the ontology and can go
> from the ontology to the lexicon. This vision obliges you to use lexical
> concepts instead in the reverse way (from lexicon to ontology). In my view,
> however, the modelling mechanism should be the same no matter you go
> bottom-up or top-down, and the "lexical entry" <-> "lexical sense" <->
> "reference" path covers both perfectly.
>
 The current axiomatization of lexical sense requires a reference. It is
possible to still have a valid resource without a reference (due to the
open world assumption) however it is a quality issue for the lexicon. We
could look to change this axiom, however we should be very conservative
about changing this, and personally I would prefer to recommend that in
cases (like bilingual dictionaries) where lexical senses are used, a dummy
reference is still introduced to promote interoperability.

> Maybe I did not get it right, but I understood that LexicalConcept was
> introduced in the last ontolex version to cover structures such as
> WordNet-like synsets (i.e., groups of words that share a common definition
> and underlying meaning), which might be suitable for some dictionaries but
> not for all of them. In other cases (specially to support translations)
> LexicalSenses  could  be enough.
>
Yes, I think we agree, but most print dictionaries will have to introduce a
lexical concept to be modeled in OntoLex

*About translation: We have previously distinguished between translation
links as a process as opposed to interlingual synonymy. Interlingual
synonymy is represented in OntoLex-Lemon by means of using the same lexical
concept or ontology reference for lexical entries in different languages,
hence translation is implicit in the multilingual lexicalizations of a
concept. The VarTrans module models translation as a process that starts
with a particular sense of a lexical entry and translates it to a sense of
another lexical entry, the link represented using the VarTrans module is
thus the result of this process.

> Best regards,
>
> Jorge
>

Regards,
John

[1] I mean something like this http://www.teanglann.ie/ga/fgb/focal
[2] Seppälä, Selja, Alan Ruttenberg, Yonatan Schreiber, and Barry Smith.
2016a. Definitions in Ontologies. Cahiers de lexicologie 2(109, La
définition):173–206.

>
>
>
>
>
> 2017-10-23 12:56 GMT+02:00 John McCrae <john@mccr.ae>:
>
> >
>
> > HI all,
> >
> > I suggested another version that follows the existing specification
> https://www.w3.org/2016/05/ontolex/
> >
> > We wrote that
> >
> > "A* lexical sense* represents the lexical meaning of a lexical entry
> when interpreted as referring to the corresponding ontology element"
> >
> > And in contrast
> >
> > "we would like to express the fact that a certain lexical entry evokes a
> certain mental concept rather than that it refers to a class with a formal
> interpretation in some model. Thus, in lemon we introduce the class*
> Lexical Concept*"
> >
> > Thus, in my interpretation, senses in traditional dictionaries should
> always be modeled as lexical concepts.
> >
> > Lexical sense has always been quite a technical concept, and honestly I
> think calling it 'lexical sense' has created much confusion (way back in
> the first Lemon model I had proposed to call it a* sememe)*. I think we
> need to make a clear and unambiguous definition of lexical sense, that
> precludes its usage without an ontological reference, as this is how it is
> defined both in the final specification and in the OWL code (*reference
> exactly 1*).
> >
> > I have created a flowchart (for discussion, attached) to try and explain
> (IMHO) the differences between senses, concepts and references.
> >
> > I also noted that there is no need to define a new example object if it
> only has an rdf:value, in this case using a single skos:example triple is
> both more compact and inter-operable.
> >
> > Regards,
> > John
> >
> > On Mon, Oct 23, 2017 at 10:21 AM, Julia Bosque Gil <jbosque@fi.upm.es>
> wrote:
>
> >>
>
> >> Hi, Philipp, all:
> >>
> >> We have a first draft of the RDF for Francesca's PLI *verre, *
> *mousse, *and *estomaquer *examples in the shared document
> <https://docs.google.com/document/d/1TogPjrLyJS0OK5pzww28751MX7179-NzCIsDdzae65o/edit>.
> Francesca, Fahad and I were working on it the other day, but some things
> are still unclear to us (marked in red color and with comments on the
> margin).
> >>
> *>> Summary:*
> >>
> >> - Lexical definitions are included at the LexicalSense level and the
> encyclopedic one at the LexicalConcept level (I seem to remember this was
> suggested during our last telco)
> >> - The decomp module is used to relate *maison de verre, petite verre,
> etc.* to* verre*.
> >> - At the end of the RDF you'll see how option 3' (from the ones we
> discussed in September), with DictionaryEntries and
> DictionaryEntryComponents, could be applied here if we wanted to record
> that both the lexical entry *verre* as well as *maison de verre, petite
> verre, *etc. belong to the same dictionary entry, and the latter are not
> considered dictionary entries themselves in the PLI.
> >>
> >> Best regards,
> >>
> >> Julia
> >>
> >>
> >> 2017-10-22 18:08 GMT+02:00 Philipp Cimiano <
> cimiano@cit-ec.uni-bielefeld.de>:
>
> >>>
>
> >>> Dear all,
> >>>
> >>>  we agreed to have our next regular ontolex skype call next Tuesday,
> >>> 24th of October at 14:00 CET.
> >>>
> >>> We will do the call by skype.
> >>>
> >>> I will not be available as I am currently attending the ISWC conference
> >>> in Vienna, but John agreed to lead the teleconference.
> >>>
> >>> The main outcome could be to provide a proposal for how to model the
> >>> Petit Larousse examples provided by Francesca during the teleconference
> >>> last week.
> >>>
> >>> Did anyone manage to have a look and try to provide some RDF code that
> >>> can be discussed during the telco next Tuesday?
> >>>
> >>> I have cleaned up a little bit the minutes from the last skype call:
> >>>
> >>> https://www.w3.org/community/ontolex/wiki/Teleconference,_20
> 17.10.10,_15-16_pm_CET
> <https://www.w3.org/community/ontolex/wiki/Teleconference,_2017.10.10,_15-16_pm_CET>
> >>>
> >>> Greetings,
> >>>
> >>> Philipp.
> >>>
> >>>
> >>> --
> >>> --
> >>> Prof. Dr. Philipp Cimiano
> >>> AG Semantic Computing
> >>> Exzellenzcluster für Cognitive Interaction Technology (CITEC)
> >>> Universität Bielefeld
> >>>
> >>> Tel: +49 521 106 12249
> >>> Fax: +49 521 106 6560
> >>> Mail: cimiano@cit-ec.uni-bielefeld.de
> >>>
> >>> Office CITEC-2.307
> >>> Universitätsstr. 21-25
> >>> 33615 Bielefeld, NRW
> >>> Germany
> >>>
> >>>
> >>
> >>
> >>
> >> --
> >>
> >> Julia Bosque Gil
> >> PhD Student
> >> Ontology Engineering Group <http://www.oeg-upm.net/>
> >> Departamento de Inteligencia Artificial
> >> Universidad Politécnica de Madrid
> >
> >
>
> --
> Jorge Gracia, PhD
> Ontology Engineering Group
> Artificial Intelligence Department
> Universidad Politécnica de Madrid
> http://jogracia.url.ph/web/
>