W3C home > Mailing lists > Public > public-ontolex@w3.org > April 2013

Re: semiotics.owl

From: Aldo Gangemi <aldo.gangemi@cnr.it>
Date: Thu, 18 Apr 2013 17:14:36 +0200
Cc: Aldo Gangemi <aldo.gangemi@cnr.it>, Guido Vetere <gvetere@it.ibm.com>, Elena Montiel Ponsoda <elemontiel@gmail.com>, public-ontolex <public-ontolex@w3.org>
Message-Id: <604FD42E-0FA0-4918-A7CF-5D08D7BC3CCB@cnr.it>
To: John McCrae <jmccrae@cit-ec.uni-bielefeld.de>
Thanks John, I think we are slowly clarifying our respective intuitions of these matters, so let's take one more step :)

On Apr 18, 2013, at 3:08:53 PM , John McCrae <jmccrae@cit-ec.uni-bielefeld.de> wrote:

> Hi,
> 
> My understanding of Sinn and Bedeutung in relation to logical theorems is that the terms of the formula are extensional and that the predicates are intensional. In the context of OWL this would mean that classes and properties are intensional and that individuals are extensional.

That's not the whole truth … besides Frege's terms, which are heavily characterised by philosophical debate, and I suggest to keep for a different thread, e.g. "history of semiotic notions", the logical issue is about the difference between intension and extension. Sorry for going back to the basics of formal logic, but we need to verify our assumptions at the very basics ;)

Formally speaking, an intension is just a constant, since anything we can say of its semantics is either intuitive or deferred to other (extensional!) interpretations, typically in terms of "possible worlds". Therefore, as soon as we populate a formal theory (or an ontology) with constants, be them typed as owl:Class, owl:ObjectProperty, owl:NamedIndividual, etc., they have an intension.

An extension is traditionally much clearer, since it is the set of individuals or tuples of individuals from the domain of interpretation that satisfy the axioms in which a constant participates (I'm intentionally informal here).

In other words, dbpedia:Person has an (intuitive) intension (formally we only have the constant "Person"), and the set of all individuals belonging to dbpedia:Person as its extension. But also dbpediaresource:Barack_Obama has an intension (its name and its intuitive or possible world interpretation, not addressed by OWL), and an extension, i.e. the actual individual that satisfies the axioms in which dbpresource:Barack_Obama participates, e.g. having certain birthdate, children, roles, etc.

Individuals are not "extensional": they have both an intension and an extension, and they also provide the extensional interpretation for classes and relations.

> In particular I think this correlates with the difference that it is informative to say 
> 
> dbpedia:Influenza owl:sameAs icd:J10
> 
> But the use of owl:equivalentClass and owl:equivalentProperty are instead to give definitions:
> 
> lexinfo:TransitiveFrame owl:equivalentClass [ ∃ subject.Argument ∏ ∃ directObject.Argument ]

Consequently, I have to disagree here. dbpedia:Influenza owl:sameAs icd:J10 is an identity relation between two individuals; if you type dbpedia:Influenza and icd:J10 as owl:Class, then you may want to say: dbpedia:Influenza owl:equivalentClass icd:J10. In other words, the interpretation of a constant is bound to its context (at least in languages that allow to do that, e.g. OWL2): when I use owl:sameAs, I'm interpreting entities as individuals, when I use owl:equivalentClass, I'm interpreting entities as classes.

(Peircean) semiotics is idle with respect to that. Being an expression, a meaning or a reference is a role, not a rigid characteristic. In fact, I can take the string "person", and say that the *expression* person may express the *meaning* wordnet-wordsense-person-1, and may denote the reference dbpedia:Person (as the extensional aspect of that owl:Class). But I can also (albeit useless it can be) say that dbpedia:Person is a capitalised expression, or that dbpedia:Person is the meaning of "person" (as the intensional aspect of that owl:Class).

Following this line of reasoning, semiotics complements formal semantics with an explicit and functional encoding of the symbol-intension-extension distinction that is only (partly) addressed by formal logic in its meta-theory. 

> 
> Furthermore, classes are clearly more Sinn than Bedeutung as it is possible to define an OWL class with no extension (i.e., that is equivalent to ⊥)

I do not see any problem in considering an empty extension, actually the zero or empty set is a very important notion, and it's an extension, even if no individual is contained in it. Classes have both an intensional and an extensional aspect, cardinality doesn't count for that.

> 
> However it is quite justified to assume that all OWL classes are in fact extensions over the set of OWL individuals, and perhaps this is the better consensus understanding.

Ok

> 
> Firstly, can I be clear that what we refer to a sense in OntoLex has generally not been 'Sinn' but as in the sense of WordNet, LMF, etc. as "a lexical sense is all usages of a given word to express a concept". This is clearly extensional and lexical. 
> 
> To clarify this, in the sentence:
> 
>     "The internet is full of pictures of cats"
> 
> 'cats' is in the extension of the lexical form for the plural of the lexeme 'cat'

Here the actual lexical form "cats" can be the individual in the extension of the lexical form "cats", quite a perverse example, but ok ;)

> 'cats' is in the extension of the lexical entry for the lexeme 'cat'

Disagree. The lexeme "cat" has an extension that can be (again perversely) the individual "cat", but not "cats", unless you want to use "cats" as its identifier. Maybe you want to say that a lexical entry is a class of lexical forms, but I doubt this is a good modelling choice. A class is not the same as an actual collection of things: just an abstract collection in the world of set theory.

> 'cats' is in the extension of the lexical sense for the usage of the lexeme 'cat' to refer to/express the species felis catus

Also disagree. A lexical sense, being intensional, is an individual, we want to promote it to an owl:Class, but anyway not to express a set of lexical senses, but of real world individuals. 
What you want to say is maybe that there also exists a collection of word senses expressed by lexical forms of the lexical entry "cat", which can be put into an equivalence class to be somehow *related* to the sense "cat".

> 
> That is the extension of a lexical sense is a set of words.

I disagree with that. We need then to distinguish between a semiotic notion of "sense" (lexical or not), and a linguistic notion of an "equivalence class of words", whose identity criterion is provided by a(n actual) semiotic sense.

> The primary reason for making this distinction is to allow for different levels of annotations:
> 
> number=plural is a property of the lexical form 'cats'

Fine

> partOfSpeech=noun is a property of the lexical entry 'cat'

Fine

> register=neutral is a property of the lexical sense 'cat'/felis catus
> 

Fine. But these three annotations do not require us to confuse semiotic/logical levels. We can have relations between forms, entries, and senses without any formal confusion.

> From this understanding I would conclude that in Aldo's model Lexical Form, Entry *and* Sense are all in fact Expressions.

Can be, not "are". As constants, they are *also* expressions, if seen as such. But a sense can be seen more typically as meaning, and anyway all of them can be seen as references, if we go "de dicto", as in the sentence "the lexical form cat is nice".

> 
> Returning to Sinn and Bedeutung, it seems clear that we could model difference as link between the lexical sense and the ontology by means of the following triangle
> 
> <cat> a ontolex:LexicalEntry ;
>    ontolex:sense <cat::01> .
> 
> <cat::01> a ontolex:LexicalSense ;
>    ontolex:expresses <Cat> ;
>    ontolex:reference <FelisCatus> . % or ontolex:denotes
> 
> <Cat> a ontolex:Meaning % subclassOf skos:Concept ;
>    ontolex:conceptualizes <FelisCatus> .
> 
> <FelisCatus> a owl:Class .
> 
> This introduces a new meaning class for the intension of Cat distinct from the extensional (ontological) definition. Does this seem reasonable?
> 

With the comments I made, we can find a way of keeping all together. Will be back soon, I have to move now.
Thanks
Aldo

> Regards,
> John
> 
> 
> 
> 
> 
> On Wed, Apr 17, 2013 at 4:06 PM, Aldo Gangemi <aldo.gangemi@cnr.it> wrote:
> Ciao Guido, 
> I am with you (and John) in pointing at the intensional import of ontology concepts, but I think treating them as extensional objects seems more practical and cleaner, specially as a default assumption. Enumerated classes (aka "nominals") are no different under this respect, except that you have an explicit complete extension, which is not the case with regular classes due to open world assumption.
> It seems that hardly a KR language allows to talk about concepts as twofold entities, i.e. with a different identity for each aspect … OWL2 punning does something, but only wrt reasoning, because the intensional and extensional aspects of a class are flattened on a same identifier.
> BTW, if in OntoLex we accept (as semiotics does) to use either sense or reference properties to talk about logical constants, we should be able to inject some limited form of duplicated interpretation, at least at the interface between ontologies and lexica.
> 
> Concerning Frege, that is the most common interpretation, but I'm not sure everyone agrees. Interestingly, the translation of Bedeutung is usually "reference", and not "referent".
> 
> Aldo
> 
> On Apr 17, 2013, at 1:45:01 PM , Guido Vetere <gvetere@it.ibm.com> wrote:
> 
>> This looks like the classic distinction of Sinn vs Bedeutung (Frege), isn't it? If the range of 'referent' has to be intended extensionally (Bedeutung), then pointing a Meaning (Sinn) to an ontology concept should be better clarified, since ontology concept may still be intensional constructs, and in fact enumerated concepts like the one in John's example are not common in average ontologies.  I would say that the class supplied as 'referent' could be either intended as an extension (like in the example) or a constraint that a suitable interpretation must satisfy. 
>> 
>> Guido Vetere
>> Manager, Center for Advanced Studies IBM Italia
>> _________________________________________________
>> Rome                                     Trento
>> Via Sciangai 53                       Via Sommarive 18
>> 00144 Roma, Italy                   38123 Povo in Trento
>> +39 (0)6 59662137                 
>> 
>> Mobile: +39 3357454658
>> _________________________________________________ 
>> 
>> 
>> Elena Montiel Ponsoda <elemontiel@gmail.com>
>> 17/04/2013 12:06
>> 
>> To
>> public-ontolex@w3.org
>> cc
>> Subject
>> Re: semiotics.owl
>> 
>> 
>> 
>> 
>> 
>> Hi Aldo, John, all,
>> 
>> After reading Aldo's previous e-mail on "Re: Senses, synsets and ontology mapping in WordNet", I am not so sure I agree with the example provided by John in the wiki page, that is why I would like to ask you for clarification.
>> 
>> Example: The term 'G8' expresses the OWL Class G8Countries and denotes the set {Canada,France,Germany,Italy,Japan,Russia,US,UK}, which is conceptualized by G8Countries
>> 
>> I quote from Aldo's previous e-mail:
>> Re (3):
>> " 
>> This is not the case when we want to link lexical or KOS meanings to typical ontologies, e.g. to myont:Vomit class. If Vomit is an OWL (or RDFS) class, its interpretation is *extensional* (a collection of things, vomiting events in the common interpretation), therefore it's fully justified to use ontolex:reference for representing this linking.
>> Therefore:
>>                 wordnet:wordsense-vomit-verb-1 wordnet:inSynset wordnet:synset-vomit-verb-1
>>                 wordnet:wordsense-vomit-verb-1 ontolex:reference myont:Vomit
>>                 wordnet:synset-vomit-verb-1 ontolex:reference myont:Vomit
>> 
>> "
>> Aldo, if I understood this correctly, the reference relation would be established between the WordNet synset or meaning and an OWL class "myont:Vomit" in an ontology. 
>> Therefore,  the expression to vomit expresses the meaning (sense or synset) corresponding to that verb
>>                    the meaning (sense or synset) conceptualizes an OWL (or RDFS) class
>> 
>> If we extrapolate this to the G8 example, we would say that the expression G8 expresses the meaning of "a group of eight of the richest industrial countries in the world...", which, in its turn, conceptualizes the OWL Class G8Countries, that contains and denotes the set of countries {Canada,France,Germany,Italy,Japan,Russia,US,UK}, as instances. 
>> 
>> I would rather agree with this view, since I my interpretation of the semiotic triangle, the "Referent" corresponds to a conceptualization or ontology of a certain world. 
>> In fact, I think that if we assume this view, the LexicalEntry-Sense-OntologyClass path has more sense. Could you agree with this view?  
>> 
>> Best,
>> Elena.
>> 
>> 
>> 
>> El 16/04/2013 20:27, Aldo Gangemi escribió: 
>> Hi, John has made a nice summary of semiotics.owl, and some useful questions. John, thanks for the analysis :) 
>> Answers below. 
>> 
>> On Apr 16, 2013, at 6:45:53 PM , John McCrae <jmccrae@cit-ec.uni-bielefeld.de> wrote: 
>> 
>> Hi Aldo,
>> 
>> I was trying to synthesize the semiotics.owl ontology you sent around. I made some notes here
>> 
>> http://greententacle.techfak.uni-bielefeld.de/~jmccrae/semiotics.owl.html
>> 
>> I had some things I didn't fully understand
>> 
>> Naming of 'reference', looking at the comments it seems to be what is (from my experience) called a 'referent' in the semiotics literature. In fact, more confusingly some authors seem to use reference for meaning. I think this has confused a lot of the OntoLex discussion as in lemon we use reference as the term for the meaning in the ontology. (Honestly this is a total accident, it was originally chosen to harmonize with LMF's 'monolingual external reference', used to cite external resources for senses). 
>> 
>> I opted for "reference" because "referent" typically bears a realistic flavor in philosphical debates. Besides that, no problem in using "referent", as in the Ogden-Richards version of the triangle. 
>> 
>> 
>> What is a manifestation, it has no annotations? 
>> 
>> Right, I should add it. A Manifestation is the material occurrence of an Expression, e.g. Dante's Comedy (Expression) can be manifested in an eBook or a paper book. The same term with similar meaning is used in FRBR vocabulary. 
>> 
>> 
>> The model has linguistic (speech?) acts, I was wondering if there were any practical examples of how to model a speech act 
>> 
>> I'll add it, and yes, it's an event (type). Example: "Talleyrand said "Si cela va sans dire, cela ira encore mieux en le disant" during the Vienna Congress in 1814." 
>> That is the report of a linguistic act (a declarative speech act that reports another, subtler speech act), which can be modeled as follows (in Turtle, and using a default namespace ":" for a domain ontology, full example at http://www.ontologydesignpatterns.org/ont/example/talleyrandquotation.ttl): 
>> 
>> :lingAct_1 a :Quote . 
>> :Quote rdfs:subClassOf semiotics:LinguisticAct . 
>> :lingAct_1 situation:isSettingFor _:agent . 
>> _:agent a agentrole:Agent . 
>> _:agent :authorOf "http://dinoutoo.pagesperso-orange.fr/histo/tal1.htm"^^xsd:anyURI . 
>> :lingAct_1 situation:isSettingFor :lingAct_2 . 
>> :lingAct_2 a :Say . 
>> :Say rdfs:subClassOf semiotics:LinguisticAct . 
>> :lingAct_2 situation:isSettingFor :Talleyrand . 
>> :Talleyrand a agentrole:Agent . 
>> :lingAct_2 situation:isSettingFor _:time . 
>> _:time :inDate "06-10-1814"^^xsd:date . 
>> :lingAct_2 situation:isSettingFor _:sentence . 
>> _:sentence a semiotics:Expression . 
>> _:sentence ontolex:lexicalForm "Si cela va sans dire, cela ira encore mieux en le disant"^^xsd:string . 
>> :lingAct_2 situation:isSettingFor _:meaning . 
>> _:meaning a semiotics:Meaning . 
>> :lingAct_2 situation:isSettingFor _:reference . 
>> _:reference a semiotics:Reference . 
>> _:reference :partOf :CongressOfVienna . 
>> 
>> if we know more about that quotation, we might add something more about _:meaning: 
>> 
>> _:meaning semiotics:relatedMeaning :Clarity . 
>> :Clarity a semiotics:Meaning . 
>> _:meaning :modality :needed . 
>> :needed a :Modality . 
>> _:meaning :inContext :InternationalTreaty . 
>> :InternationalTreaty a dbpedia:Event . 
>> 
>> 
>> Some unreferenced elements: agent, hasComponent... what is their purpose exactly? 
>> 
>> Those are actually references, but in other - imported - patterns (agent role, situation, cpannotationschema) 
>> 
>> 
>> Regards,
>> John 
>> 
>> Aldo 
>> 
>> 
>> -- 
>> Elena Montiel-Ponsoda
>> Ontology Engineering Group (OEG)
>> Departamento de Inteligencia Artificial
>> Facultad de Informática
>> Campus de Montegancedo s/n
>> Boadilla del Monte-28660 Madrid, España
>> www.oeg-upm.net
>> Tel. (+34) 91 336 36 70
>> Fax  (+34) 91 352 48 19 
>> 
>> 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)
> 
> 


Received on Thursday, 18 April 2013 15:15:07 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 16:36:30 UTC