Re: Meaning and Semiotics - Issues for Modelling

Hi all,

I think this level proposed by Piek is interesting, I will try to give my
answers

1. Is there any semantic aspect of a word sense (I prefer lexical unit)
that cannot be represented in an ontological model?

This primarily depends on the grounding of the ontology. For example, we
may define "big" as "having a high size value for its class" but this
requires an ontology language capable of representing "having a high value"
and a perception of "size". As such ultimately all ontologies require some
basic predicates that can't be derived within the ontology but are derived
externally (e.g., "size" by a computer vision system). From the point of
the lexicon then it doesn't matter much whether the symbol in the ontology
is further defined (e.g., "big") or a basic predicate (e.g., "size"), that
is the lexicon should treat "big" as just another symbol in the ontology.
This is even more clear if you consider the set of basic symbols as
essentially arbitrary, in which case there is no reason to assume "big" is
not itself a basic symbol (e.g., the output of some system).

2. Do you want to model any semantic aspect that characterizes a word sense
also in the ontology?

Again, this is mostly a matter of symbols: "buy" and "sell" are likely
simply inverse properties. The modelling question is really then whether we
allow modelling such as

   1. Inverse lexical mapping: "buy"@en meansInverseOf Sell
   2. Mapping to anonymous classes: "buy"@en means [ a Property ; inverseOf
   Sell ] .
   3. Mapping (only) to named classes: "buy"@en means Buy . Buy a Property
   ; inverseOf Sell .

I strongly prefere 2 and 3, as it avoids reasoning in the lexical layer
(i.e., the OntoLex group does not need to define new axioms for reasoners
(e.g., Pellet))

3. What do we do with the situations that lexicons are far more richer than
any ontology available and thus we cannot provide sufficient ontological
labels to model the lexicons.

Similar answer to above: extend the ontology, with either anonymous or
named new concepts

Regards,
John

On Fri, Aug 10, 2012 at 10:05 AM, Piek Vossen <piek.vossen@vu.nl> wrote:

> Dear all,
>
> I would like to discuss this at another level. We should first answer the
> question:
>
> 1. Is there any semantic aspect of a word sense (I prefer lexical unit)
> that cannot be represented in an ontological model?
>
> It may not be easy but I think you can, if you allow semantics in the
> ontology that incorporates probabilities and prototypicality.
> I think that any formalization of lexical meaning can be turned into an
> ontological meaning, simply because it is a formalization.
> if it is not a formalization then the lexical meaning is ill-defined and
> we need to do more (empirical) work to learn about the word and its usage.
>
>
> 2. Do you want to model any semantic aspect that characterizes a word
> sense also in the ontology?
>
> This is another question. If we want to model pure logical reasoning,
> there may be many lexical aspects (not just the pragmatic knowledge) that
> we do not need
> in the ontology. We do not need to represent “buy” and “sell” separately
> to reason over de financial transaction process.
>
>
> 3. What do we do with the situations that lexicons are far more richer
> than any ontology available and thus we cannot provide sufficient
> ontological labels to model the lexicons.
>
> This is a more practical and pragmatic question. If the lexicon is so
> large, complex and rich, why not use a two-layered solution where lexical
> relations take the burden off the ontology and the ontology takes the
> burden of deeper reasoning (need to define how deep we need to go). So in
> the lexicon, I can say that one word is the informal word for “eat” and
> another word is the neutral label for “eat”. In the ontology, we just have
> “eat”. Many lexicalized concepts are either pragmatic variants or can be
> defined using intersecting
> properties as described by Philipp for “bald”.
>
> cheers
>
> Piek
>
>
> On Aug 10, 2012, at 9:41 AM, Guido Vetere wrote:
>
> > Philipp Cimiano <cimiano@cit-ec.uni-bielefeld.de>
> > 09/08/2012 21.56
> >
> > To
> >
> > public-ontolex@w3.org
> >
> > cc
> >
> > Subject
> >
> > Re: Meaning and Semiotics - Issues for Modelling
> >
> > Dear John, Guido,
> >
> >  thanks for getting this discussion started.
> >
> > @Guido: I also agree that we need senses, but not for the same
> > argument that you give. I think that polysemy could be represented
> > without the type of object that John calls a "Sense".
> >
> > Take your example "bald" which according to WordNet can mean (among
> > other things):
> >
> > 1) bald, barefaced (with no effort to conceal) "a barefaced lie"
> > 2) bald, bald-headed, bald-pated (lacking hair on all or most of the
> scalp)
> > "a bald pate"; "a bald-headed gentleman"
> >
> > This could be modelled without a mediating sense as follows:
> >
> > Lexical Entry: bald (adjective) ----- mapsTo ----- Concept: obvious
> >                                              .......... mapsTo -----
> > Concept: hasNoHair
> >
>
> Philipp, you are right, I should have been more accurate. Working with a
> legacy dictionary led us to take 'meaning objects' for granted. Actually,
> if you look at standard lexicography, you can see that reifying the
> association between lexical entries and concepts (whatever they are) comes
> from the need of predicating on these relations. For instance, many
> 'senses' come with specific grammatical constraints, e.g. for nouns, plural
> is often used to mean something different from a mere collection of
> individuals, as in Italian 'acqua' (water) and 'acque' (thermal
> treatments). Many dictionaries (including Senso Comune) also have usage
> marks, to tell whether a sense (note: not a lexeme) is (presumably) part of
> the core linguistic competences, or is literary, technical, etc. Note again
> that this information cannot be attached to lexical entries, unless you
> don't want to overload homography.
>
> > And coming to your modelling, what exactly is the difference in your
> > view between mapping to a Non-Physical Subclass "Meaning" compared
> > to mapping directly to a symbol representing a concept?
> >
> > For example, if in a given ontology I have the concepts  onto:Human
> > and onto:hasNoHair, I could create the intersection
> > onto:bald = (onto:Human AND onto:hasNoHair) and then specify the
> following:
> >
> > Lexical Entry: bald (adjective)   ---- has Sense --- bald_adjective-1
> >
> > bald_adjective_1  ---- has Def --- "having a scalp wholly or partly
> > lacking hair"@en
> >
> > bald_adjective_1 ---- hasMeaning ----  onto:bald
> >
> > With this I would have also modelled the "Characterizes ony --->
> > Human part of your representation".
> >
> > However, it seems to me that this only works for intersective
> > adjectives in this way.
> >
> > Let's take the case of the German verbs "fressen" and "essen".
> >
> > They both map to the same concept "TakingInSolidFood" (Wordnet's
> > first sense for "eat"), but they differ in their selectional
> restrictions.
> > In the case of "fressen", the agent is implied to be an animal,
> > while in the case of "eat", the agent is supposed to be a human. It
> > seems to me that this can not be modelled by mapping to one single
> > concept as in the case of the intersective adjective "bald".
> >
> > Interestingly, WordNet defines two different senses for "eat" in
> > English, one corresponding to eating animals (with synonym "feed")
> > and one corresponding to eating humans.
> >
> > And a point concering "vagueness":  using the direct mapping
> > approach, you also do not commit to the fact that "bald things" exist.
> >
>
> Separating linguistic senses from ontological categories is at the core of
> our project Senso Comune. This is very crucial and would deserve a longer
> discussion.
> To start, note that your definition of onto:bald sets bald to be one with
> no hairs. Actually, you know that things are different: I'm usually told to
> be bald, but I still have some hair. I wonder when I've lost the hair that
> caused me to be told bald for the first time, but I don't find an answer
> :-) Seriously, in contrast to nowadays (computer science) 'ontologies',
> that take vagueness very easy (e.g. they set 'bald' as a primitive concept
> and leave people to provide the interpretation, like any terminology), we
> looked at a model in which ontology is limited to a minimal set of
> categories, for which, in principle, we could provide identity criteria (or
> at least we can try). Our interdisciplinary team appears to be comfortable
> with saying that 'bald' is just an ideal, social notion, a 'flatus vocis'
> that (vaguely) characterize 'human beings' (category). In our work, this
> has been modeled by setting that the only 'ontological type' of a linguitic
> unit is a social object called 'meaning', and that meanings characterize
> things that belong to 'a-priori' ontological categories. This way,
> vagueness is clearly framed into human semantics, i.e. the way in which
> humans use words. Of course, here we have to deal with legacy ontologies,
> which are far from being well-founded sets of categories, but I think that
> this discussion can enlighten some aspect of our work.
>
>
> > Btw. just for interest: Where does the term "acceptation" actually come
> from?
>
> Comes from Italian 'accezione', which is 'sense' in the lexicographic
> jargon. We are aware that in English the term is quite uncommon.
>
> >
> > Cheers,
> >
> > Philipp.
> >
> > Am 08.08.12 12:42, schrieb Guido Vetere:
> > John,
> >
> > We do need senses, since lexical units (lexemes) may have different
> > meanings (polysemy). Whether to call them 'sense', 'acceptation',
> > 'sememe', 'meaning', etc, depends (in my opinion) on how we want to
> > qualify these objects with respect to a comprehensive model (theory).
> >
> > In Senso Comune, for instance, we called them 'acceptation' (but we
> > are currently looking for a more comfortable word) because we
> > started from representing lexicographical notions, i.e.
> > (ontologically) Information Objects which are constitutive of a
> > special kind of books, namely Dictionaries. Moreover, our senses
> > (acceptations) are not directly mapped to classes representing 'real
> > things', but with (sub)classes of a Non-Physical Entity called
> > Meaning. Possibly, these classes are defined by specifying the kind
> > of real thing they characterize, which is, ultimately, the
> > 'ontological commitment' behind the word (à la Quine).
> >
> > To give an example, the adjective 'bald' in the sense of 'lacking
> > hairs') may look like this:
> >
> > Lexical Entry: bald (adjective)
> > Meaning
> > Thing
> > |__ part -> Sense 1:  Definition "having a scalp wholly or partly
> > lacking hair"     ------ maps to --->  |__ isa -> Meaning_of_bald_1
> > ---- characterizes-only ---->   |__ isa -> Human
> >
> > Note that, this way, we can account for linguistic vagueness, since
> > we don't commit ourselves to say that, in the World, there are 'bald
> > things'  - even though, if you look at me or Aldo, you wouldn't have
> > doubts about that :-)
> >
> > Regards,
> >
> > Guido Vetere
> > Manager, Center for Advanced Studies IBM Italia
> > _________________________________________________
> > Rome                                     Trento
> > Via Sciangai 53                       Via Sommarive 18
> > 00144 Roma, Italy                   38123 Povo in Trento, Italy
> > +39 (0)6 59662137                 +39 (0)461 312312
> >
> > Mobile: +39 3357454658
> > _________________________________________________
> >
>
> >
> > John McCrae <jmccrae@cit-ec.uni-bielefeld.de>
> > Sent by: johnmccrae@gmail.com
> > 06/08/2012 14.54
> >
> > To
> >
> > public-ontolex <public-ontolex@w3.org>
> >
> > cc
> >
> > Subject
> >
> > Meaning and Semiotics - Issues for Modelling
> >
> >
> >
> >
> > Hi all,
> >
> > I thought I would get the ball rolling on the requirements "Express
> > Meaning with respect to ontology". I fleshed out two of the key
> > issues for modelling the lexicon-ontology interface here
> >
> > http://www.w3.org/community/ontolex/wiki/
> > Specification_of_Requirements/Lexicon-Ontology-Mapping
> >
> > They are
> > Do we need a lexical sense/acceptation object between the lexical
> > entry and the ontology entity?
> > How do we interpret different lexical classes relative to an ontology
> > Does anyone have comments?
> >
> > Regards,
> > John
> >
> >
> > IBM Italia S.p.A.
> > Sede Legale: Circonvallazione Idroscalo - 20090 Segrate (MI)
> > Cap. Soc. euro 347.256.998,80
> > C. F. e Reg. Imprese MI 01442240030 - Partita IVA 10914660153
> > Società con unico azionista
> > Società soggetta all’attività di direzione e coordinamento di
> > International Business Machines Corporation
> >
> > (Salvo che sia diversamente indicato sopra / Unless stated otherwise
> above)
> >
>
> > --
> > Prof. Dr. Philipp Cimiano
> > Semantic Computing Group
> > Excellence Cluster - Cognitive Interaction Technology (CITEC)
> > University of Bielefeld
> >
> > Phone: +49 521 106 12249
> > Fax: +49 521 106 12412
> > Mail: cimiano@cit-ec.uni-bielefeld.de
> >
> > Room H-127
> > Morgenbreede 39
> > 33615 Bielefeld
> IBM Italia S.p.A.
> Sede Legale: Circonvallazione Idroscalo - 20090 Segrate (MI)
> Cap. Soc. euro 347.256.998,80
> C. F. e Reg. Imprese MI 01442240030 - Partita IVA 10914660153
> Società con unico azionista
> Società soggetta all’attività di direzione e coordinamento di
> International Business Machines Corporation
>
> (Salvo che sia diversamente indicato sopra / Unless stated otherwise above)
>
>
> Piek Vossen
> Professor Computational Lexicology
>
>
>
>
> T +31 (0)20 59 86457 *|*  piek.vossen@vu.nl* | *http://www.vossen.info *|*
> ADDRESS: de Boelelaan 1105, 1081 HV Amsterdam, The Netherlands *| *
> Disclaimer<http://www.vu.nl/nl/over-de-vu/vu-website/e-mail-disclaimer/disclaimer-tekst-e-mail/index.asp>
>
>
>
>
>
>