Re: Summary of Telco 07.02.2014

Yes, I am not sure "normal people" understand what is meant by
"subsumption" or "shared superconcept" either!

Gilles' example is essentially the same as the translation "gehen"=>"to go"
from German to English

I can say that "ich gehe zum Post"/"I go to the post office" but most
Germans look at me funny when I say "ich gehe nach Rußland" (for
non-Germans this implies I am walking to Russia)

As such we can say that "gehen" has reference "MovementByWalking" and "to
go" has reference "Movement". If we take the definition of translation as
it stands, this means that "gehen" is not a translation of "to go". The
only solution would be to create an extra sense of "to go" in English that
refers to "MovementByWalking", which is an artificial solution from the
point of view in English.

I don't think it is a good solution either to say that "gehen" and "to go"
are cultural equivalents for obvious reasons.

Regards,
John


On Thu, Feb 20, 2014 at 10:27 PM, Philipp Cimiano <
cimiano@cit-ec.uni-bielefeld.de> wrote:

>  John,
>
>  I know that "directly ontologically related either through subsumption or
> via a shared superconcept" is sort of meaningless, but normal people that
> do not understand OWL understand; in some sense this refers to an
> explicitly materialized concept hierarchy and directly means then: direct
> super or superconcept.
>
> In any case, I am fine with changing the definitions, that's what we
> should discuss in our telco tomorrow.
>
> But a question in return: it is not clear to me how we formally
> distinguish the cases of "translation" and "cultural equivalent".
>
> We should IMHO distinguish a lexical substituion that has no change form
> an ontological point of view and a lexical substituion that implies a
> change from an ontological point of view.
>
> Best,
>
> Philipp.
>
> Am 20.02.14 22:13, schrieb John P. McCrae:
>
> Well... we shouldn't say this.
>
>  We also shouldn't be saying things like "directly ontologically related
> either through subsumption or via a shared superconcept." (which is
> meaningless, as everything has a shared superconcept owl:Thing).
>
>  The reason is as follows: "lexical" translation is a "lexical" property
> defined in the "lexicon", as such the definitions should be explainable in
> lexical. Or put another way, if we wish to say two word senses have the
> same reference, we can do that giving them the same reference! The
> translation property indicates that there is a lexical substitution
> possible in the translation process.
>
>  We should also be more careful about implications on the ontology. In
> particular, we wrote a paper about this, and it is very clear from that
> paper that equivalence of sense can not imply equivalence of references:
> http://www.lemon-model.net/papers/senses.pdf
>
>  Regards,
> John
>
>
> On Thu, Feb 20, 2014 at 9:58 PM, Philipp Cimiano <
> cimiano@cit-ec.uni-bielefeld.de> wrote:
>
>>  John, all,
>>
>>    The thing is that we define "Translation" as implying same references.
>> That has an implication, although we do not model this axiomatically.
>>
>> Are we on the same line?
>>
>> Philipp.
>>
>> Am 20.02.14 21:55, schrieb John P. McCrae:
>>
>> Hi
>>
>>  I don't think we should be able to infer any ontological relation
>> especially not owl:sameAs from a translation property. The translation
>> property simply means that you could consider one of the senses to "align"
>> with the foreign term in translation (with the definition of align being
>> fuzzy). We merely distinguish between translations and cultural-equivalent,
>> which describe whether there is an intentional change in the meaning of a
>> sentence. If you want to say that there is ontological equivalence that
>> should be done in the ontology only.
>>
>>  To answer Philipp's earlier question, some interesting examples of
>> cognates are "Tisch"@de (=table) and "discus"@la (=disc) or "mandibola"@it
>> (=jaw) and "mandibles"@en
>>
>>  Regards,
>> John
>>
>>
>> On Thu, Feb 20, 2014 at 9:39 PM, Philipp Cimiano <
>> cimiano@cit-ec.uni-bielefeld.de> wrote:
>>
>>>  Dear Gilles,
>>>
>>> yes you are right. I understand that this might negatively affect the
>>> usability of the model. And I am fully aware that people might use it
>>> sloppily, actually without to many unwanted implications as long as you do
>>> not use an OWL reasoner ;-)
>>>
>>> The alternative would be to give up the distinction between
>>> semantics-preserving interlingual variants (translation) and
>>> non-semantics-preserving interlingual variants (cultural equivalents), or
>>> call them differently and make translation the superclass.
>>>
>>> Any opinions on this? I tend to see "translation" indeed as
>>> semantics-preserving, but this is only a gut feeling and not well-founded.
>>>
>>> Another issue: we discussed having a third type of Interlingual variant,
>>> something like a "cross-lingual paraphrase" class for the case in which
>>> "paella" is paraphrased in English as "typical rice dish from Spanish
>>> origin".
>>>
>>> Elena: I think this is what you had in mind, right?
>>>
>>> Regards,
>>>
>>> Philipp.
>>>
>>> Am 20.02.14 21:31, schrieb Gilles Sérasset:
>>>
>>> Dear Philipp,
>>>
>>>  Thanks, this clarifies the matter.
>>>
>>>  Tough I fear that the encoding of legacy lexica in ontolex will not be
>>> very easy, as most of the time, such lexica does not really make the
>>> difference... I fear many ontolex encoded lexica will use the Translation
>>> relation regardless of the implications when linked to an ontology.
>>>
>>>  Regards,
>>>
>>>  Gilles,
>>>
>>>  On 20 févr. 2014, at 20:40, Philipp Cimiano <
>>> cimiano@cit-ec.uni-bielefeld.de> wrote:
>>>
>>>  Dear Gilles,
>>>
>>> thanks for your comment. Yes indeed, if we use "Translation" in the case
>>> you mention, we would infer that concept:rice owl:sameAs
>>> concept:cooked_rice.
>>>
>>> We would infer equality of punned individuals. Technically, it does not
>>> follow though that the concepts are equivalent. It is a delicate OWL2
>>> issue.
>>>
>>> In any case, your statement is correct. If that is not as intended in
>>> your example (which I assume) then in your case the relation
>>> "CulturalEquivalent" should be used which is supposed to be used in exactly
>>> such a case where there is some direct ontological relation between both
>>> concepts, in our case concept:rice subsumes concept:cooked_rice.
>>>
>>> So the use of "Translation" is wrong in your case because it has
>>> unwanted implications.
>>>
>>> Do you agree?
>>>
>>> Philipp.
>>>
>>> Am 20.02.14 10:02, schrieb Gilles Sérasset:
>>>
>>>  Dear all,
>>>
>>>  I have a question regarding Translation.
>>>
>>>  Lets take the japanese of 御飯 (gohan), which has a meaning of "cooked
>>> rice".
>>> Lets take the english term "rice" (which refers to cooked or uncooked
>>> rice, indistinctly.
>>>
>>>  I consider both terms as translations of each others, even if they do
>>> not share the reference.
>>>
>>>  Indeed, I do think that the Translation relation is useful, as the
>>> lexicon should exist even if no conceptualization is available. It is also
>>> really useful to encode existing lexica.
>>>
>>>  But with this definition, if my lexicon state that gohan is a
>>> translation of rice, then we would legitimately infer that concept:rice
>>> owl:sameAs concept:cooked_rice.
>>>
>>>  Isn't it a problem in itself ?
>>>
>>>  Regards,
>>>
>>>  Gilles,
>>>
>>>
>>>
>>>  On 13 févr. 2014, at 16:22, Philipp Cimiano <
>>> cimiano@cit-ec.uni-bielefeld.de> wrote:
>>>
>>>    Hi Elena,
>>>
>>>    just to clarify intuitions. I am calling a Translation something
>>> which preserves the reference (no matter if literal or not).
>>>
>>> So according to what I have now it holds that:
>>>
>>>  Class: var:Translation
>>>
>>> SubclassOf:
>>>         ontolex:InterlingualVariant
>>>         ontolex:TermVariant
>>>
>>> rdfs:comment "The relation between two lexical senses in different languages the references of which are the same."@en
>>>
>>>
>>> So this means that Translation is a relation between two Lexical Senses
>>> in different languages the reference of which is the same.
>>>
>>> On the other hand, CulturalEquivalent (or simply Equivalent!) is defined
>>> as follows:
>>>
>>>  Class: var:CulturalEquivalent
>>>
>>> SubclassOf:
>>>         ontolex:InterlingualVariant
>>>         ontolex:SemanticVariant
>>>
>>> rdfs:comment "The relation between two lexical senses in different languages the references of which are directly ontologically related either through subsumption or via a shared superconcept."@en
>>>
>>>
>>>
>>>
>>> i.e. the references are directly ontologically related, does this make
>>> sense?
>>>
>>> Philipp.
>>>
>>>
>>> Am 13.02.14 16:10, schrieb Philipp Cimiano:
>>>
>>> Hi Elena,
>>>
>>> see below
>>>
>>> Am 13.02.14 13:13, schrieb Elena Montiel Ponsoda:
>>>
>>> Dear Philipp,
>>>
>>> Thanks for the updates.
>>> I have direclty modified the text in the specification (maybe I should
>>> not?), but we can still reconsider this...
>>> On the one hand, I thought it is important to specify already at the
>>> introduction that there is one type of variation that is established
>>> between LexicalEntries (i.e., define LexicalVariants), how do you see it?
>>>
>>>
>>> Yes fine, I should remove the restriction from Variants that requires
>>> LexicalSense, I will do it now.
>>>
>>> On the other, I was not so happy with the "terminology" used when
>>> dealing with cross-lingual variants, specifically when stating that
>>> Translations are literal translations...
>>>
>>> Fair enough, if the idea is removing "literal" I am agnostic ;-)
>>>
>>>  From the Translation discipline perspective, this would be problematic,
>>> IMHO.
>>>
>>>    - I think we should refer to them as Translations or Interlingual
>>>    variants (in general). That is what people interested in multilinguality
>>>    will be looking for, I think. If you think that the MultiWordNet community
>>>    would be happier with Inter-lingual variant is fine, but the translation or
>>>    terminology community will be looking for "translation". Would it be
>>>    feasible to keep both denominations? Since this is a lexicon model (for
>>>    ontologies, of course, but still we are at the lexical level), I would be
>>>    inclined to think that the most appropriate term is translation, but I am
>>>    open to change my mind... :)
>>>
>>>
>>> OK, so why not having "InterlingualVariant" as a subClass of
>>> "TermVariant" (instead of TerminologyVariant) and then
>>> Translation and CulturalEquivalent and "CulturalParaphase" as subclasses
>>> of InterlingualVariant.
>>>
>>> Would that be appropriate?
>>>
>>>
>>>
>>>
>>>    - As for the types of translation we may account for, I would talk
>>>    of "equivalents", but not identify "translations" exclusivly and explicitly
>>>    with "literal translations". I was trying to make this clear during out
>>>    last telco, but maybe I failed... :) That is why I was proposing direct
>>>    equivalents, to distinguish them from cultural equivalents.
>>>
>>>  Fair enough, if you are arguing for dropping the "literal" I am fine.
>>>
>>>
>>>
>>> As for the question in your e-mail referring to "paraphrase", yes, I
>>> think we could put it that way...
>>> Best,
>>> Elena
>>>
>>> El 13/02/2014 10:02, Philipp Cimiano escribió:
>>>
>>> Hi Elena, all,
>>>
>>>  I have updated the wiki reflecting the discussion of last week;
>>> however, I have not introduced SenseRelations explicitly yet. I am not sure
>>> we should.
>>>
>>> In any case, we agree in principle on the categories mentioned by you
>>> Elena, but I have one question on the lexical equivalent: this is
>>> essentially a paraphrase, right?
>>>
>>> Philipp.
>>>
>>> Am 07.02.14 17:27, schrieb Elena Montiel Ponsoda:
>>>
>>> Dear John,
>>>
>>> Thanks for the summary (Philipp, do not stay away... we missed you...
>>> ;)).
>>>
>>> Regarding the Translation part, I think we had a nice discussion, but we
>>> need to work a little bit more on that.
>>> I tend to think of Term Variants as within the same language
>>> (intra-lingua), and Translations between languages (inter-lingua). For this
>>> reason, I am not so sure I would like to consider Translation a Term
>>> Variant, but I will further think about it... :)
>>>
>>> In a paper we at UPM just got accepted at the LREC conference, we were
>>> proposing 3 different types of *translation equivalents*.
>>>
>>>    1. *direct equivalent *(what people normally understad as "pure
>>>    translation"): The two terms describe semantically equivalent entities that
>>>    refer to entities that exist in both cultures and languages. E.g. surrogate
>>>    mother, madre de alquiler, mère porteuse. It is true that they could
>>>    further be considered *dimensional variants*, since each
>>>    language/culture emphasizes a different aspect of the concept.
>>>    2. *cultural equivalent*: Typically, the two terms describe entities
>>>    that are not semantically but pragmatically equivalent, since they describe
>>>    similar situations in different cultures and languages. E.g., “Ecole
>>>    Normal” (FR) “Teachers college” (EN). The Prime Minister and Busdeskanzler
>>>    example would also be valid here. And I think this is the type of *link
>>>    or cross-lingual alignment you would use in ** Interlingual Indexes
>>>    for WordNets when no "direct equivalent" in available*.
>>>     3. *lexical equivalent*: It is said of those terms in different
>>>    languages that usually point to the same entity, but one of the verbalizes
>>>    the original term by using target language words. E.g., “Ecole Normal” (FR)
>>>    “(French) Normal School” (EN). The concept of Normal School does not exist
>>>    in England, but English people have verbalized it in English.
>>>
>>>
>>> Does it make sense?
>>> We will also work on this and update the wiki with examples/code
>>> accordingly.
>>> Have a nice weekend!
>>> Elena.
>>>
>>> El 07/02/2014 16:59, Philipp Cimiano escribió:
>>>
>>> Dear all,
>>>
>>> very nice, it seems that the telco was very productive without me, I
>>> should consider staying away now and then ;-)
>>>
>>> I will work this into the current document next week.
>>>
>>> Best regards,
>>>
>>> Philipp.
>>>
>>> Am 07.02.14 16:29, schrieb John P. McCrae:
>>>
>>> Hi all,
>>>
>>>  So today at the telco we had myself, Paul, Francesca, Elena and Lupe.
>>>
>>>  We discussed based on Philipp's proposal
>>>
>>>  I propose we go with the following four variants + translation:
>>>> 1) FormVariant: Relation between two forms of one lexical entry
>>>> 2) LexicalVariant: Relation between two lexical entries that are
>>>> related by some well-defined string-operation (e.g. creating an initialism
>>>> like in FAO)
>>>> 3) TerminlogicalVariant: Relation between two lexical senses (with the
>>>> same reference) of two lexical entries; the lexical entries are thus
>>>> uniquely determined; the senses might have different contextual and
>>>> pragmatic conditions (register, etc.)
>>>> 4) SemanticVariant: As 3) Relation between senses with references that
>>>> are ontologically related, either by subsumption or are children of a
>>>> common superconcept (see my paella and risotto example)
>>>> 5) Translation: As with 3), but involving entries from different
>>>> languages.
>>>> So we would have one relation between forms (FormVariant), one relation
>>>> between lexical entries (LexicalVariant), and three relations at the sense
>>>> level (TerminologicalVariant, SemanticVariant and Translation).
>>>> We might think about introducing a SenseRelation as a superclass of
>>>> TerminologicalVariant, SemanticVariant and Translation. Hypernym and
>>>> Hyponym would also be a SenseRelation in this sense.
>>>
>>>
>>>  The discussion was as follows:
>>>
>>>  *Form variants*: We discussed the need to distinguish form
>>> (inflectional) variants as opposed to lexical (entry) variants. The primary
>>> reason for this was to separate variation between LexicalEntrys and Form
>>> (as defined in the core). It was felt that the distinction between form and
>>> lexical variant was too fine-grained and that the modelling of this as
>>> variants is probably not appropriate. For example, if we consider
>>>
>>>  :Cat a LexicalEntry
>>>   ontolex:canonicalForm :Cat#CanonicalForm (writtenRep "cat"@eng),
>>>   ontolex:otherForm :Cat#PluralForm  (writtenRep "cats"@eng) .
>>>
>>>  Then modelling the relationship as
>>>
>>>  :Cat#CanonicalForm ontolex:plural :Cat#PluralForm
>>>
>>>  is inferior to (especially in the case that there are large number of
>>> inflections of a single lemma, such as an Italian verb)
>>>
>>>  :Cat#CanonicalForm ontolex:number ontolex:singular .
>>> :Cat#PluralForm ontolex:number ontolex:plural .
>>>
>>>  For these reasons, it was preferred not to introduce form variants
>>>
>>>  *Term(inological)Variants/SemanticVariant: *We agreed with the idea of
>>> introducing a superclass SenseRelation subsuming both TermVariants and
>>> SemanticVariants as follows
>>>
>>>    - TermVariants have the same reference (e.g., diachronic, diatopic
>>>    etc.)
>>>    - SemanticVariants have different references (e.g., antonymy,
>>>    "similar", (maybe?) hypernymy)
>>>
>>> It was also suggested to shorten the name TerminologicalVariant to
>>> TermVariant
>>>
>>>  *Translation: *We discussed the idea of distinguishing between
>>> (Lemma/Term) *Translation* and *Culturally-Equivalent Translation *by
>>> saying *Translation * is a *TermVariant * and *Culturally-Equivalent
>>> Translation* is a *Semantic Variant.*
>>> It was suggested that we consider introducing a class
>>> *MultilingualVariant** subsuming *Translation *and* C.E.T. *and
>>> subsumed by *SenseRelation, *for relations between languages, this
>>> would also include broader/narrower cross-lingual alignments as used in
>>> Interlingual Indexes for WordNets etc.
>>> * or cross-lingual variant or inter-lingual variant
>>>
>>>  I attach a diagram to show the proposal
>>>
>>>  Regards,
>>> John
>>>
>>>
>>>
>>>
>>>
>>>
>>> --
>>>
>>> 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
>>> Escuela Técnica Superior de Ingenieros Informáticos
>>> 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
>>>
>>>
>>> --
>>> Elena Montiel-Ponsoda
>>> Ontology Engineering Group (OEG)
>>> Departamento de Inteligencia Artificial
>>> Escuela Técnica Superior de Ingenieros Informáticos
>>> 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
>>>
>>>
>>>
>>> --
>>>
>>> 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
>>>
>>>
>>>   --
>>>  Gilles Sérasset
>>> GETALP-LIG                         BP 53 - F-38041 Grenoble Cedex 9
>>> Phone: +33 4 76 51 43 80 <%2B33%204%2076%2051%2043%2080>
>>>     Fax:   +33 4 76 63 56 86
>>>
>>>
>>>
>>>  --
>>>
>>> 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
>>>
>>>
>>>  --
>>>  Gilles Sérasset
>>> GETALP-LIG                         BP 53 - F-38041 Grenoble Cedex 9
>>> Phone: +33 4 76 51 43 80 <%2B33%204%2076%2051%2043%2080>
>>>     Fax:   +33 4 76 63 56 86 <%2B33%C2%A04%2076%2063%2056%2086>
>>>
>>>
>>>
>>> --
>>>
>>> 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
>>>
>>>
>>
>>
>> --
>>
>> 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
>>
>>
>
>
> --
>
> 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
>
>