- From: Philipp Cimiano <cimiano@cit-ec.uni-bielefeld.de>
- Date: Fri, 18 Jul 2014 14:01:58 +0200
- To: public-ontolex@w3.org
- Message-ID: <53C90CB6.1030409@cit-ec.uni-bielefeld.de>
John, what you propose is not far from my own understanding. I agree that semantic frames are in a 1:1 relationship with sense. That is exactly the intuition that I want to capture. I think my proposal to make SemanticFrame a subclass of LexicalSense was misleading and not really to the point. So to keep things really simple: I would simply propose two changes to the model: 1) add a class SemanticFrame with exactly the understanding that there is a one-to-one relation between LexicalSenses and SemanticFrames 2) My intuition is that something that has semanticArguments has to be a SemanticFrame, so I would simply specify the domain of semArg to be "SemanticFrame". The behaviour of this is that objects of type "Lexical Sense" that are claimed to have semantic Arguments are automatically infered to be "SemanticFrames"s. With that we make no explicit statement about the ontological relation between Lexical Senses and Semantic Frames other than that there are objects that can be both ;-) Regards, Philipp. Am 18.07.14 11:51, schrieb John P. McCrae: > Hi Philipp, > > I think I would agree with Armando that the semantic frame is not > /conceptually/ a lexical sense, and modelling semantic frames as > subclasses of lexical senses is conceptually unsound and unusual. > However, as we have exactly one reference of a lexical sense, it > follows that we have exactly one semantic frame, that is semantic > frames are in a 1:1 relationship with lexical senses. > > In Monnet /lemon/ we had a principle of conciseness, and thus as the > semantic frame is in a 1:1 relationship with lexical senses, it > follows that it is not necessary to assign a different URI to the > semantic frame, i.e., we expand the conceptual definition of a lexical > sense to includes its semantic frame, and eliminate the SemanticFrame > class from the model. > > In OntoLex we have a number of equivalent representations of the same > thing, so one potential solution is to create a class SemanticFrame > and duplicate the semArg properties with appropriate property chains, > e.g., > > ObjectPropety:*semFrame* (Domain: LexicalSense, Range: *SemanticFrame*) > semArg ⊒ semFrame o *semArgument* > subjOfProp ⊒ semFrame o *semSubject* > objOfProp ⊒ semFrame o *semObject* > isA ⊒ semFrame o *semMember* > * > * > This would allow users to use "semantic frames" if needed while still > remaining conceptually sound. Of course it also increases the size of > the model for no good technical reason. > > Regards, > John > > > On Fri, Jul 18, 2014 at 11:05 AM, Philipp Cimiano > <cimiano@cit-ec.uni-bielefeld.de > <mailto:cimiano@cit-ec.uni-bielefeld.de>> wrote: > > John, Armando, all, > > sorry for my late reply on this issue with the "Semantic Frame". > > I still think that it is a good idea to introduce Semantic Frame > as a subclass of "Lexical Sense". Let me try to argue a bit more: > > 1) Of course, the semantics is in the ontology, but as we all know > frames are not explicit in languages such as OWL / RDF, so the > "Semantic Frame" class would essentially stand proxy for a > structure that can be represented in terms of ontology predicates. > Imagine I have a class "GoodExchange" and a property "Lender" and > a property "borrower". Then the semantic frame associated to the > expression "X borrowed Y from Z" is represented by a particular > set of properties in the ontology, i.e. the binary properties > "lender" and "borrower". The Semantic Frame is a prox object in > the lexicon that binds these properties into a unit (gestalt) that > expresses the meaning of a syntactic frame such as "X borrowed Y > from Z". I agree this is in principle only syntactic sugar as > this can already be represented by the current vocabulary we have. > The main difference is that it makes the fact that at the ontology > side we actually have a frame with arguments more explicit and > clearer, particulary considering the following point 2: > > 2) The main reason why I am arguing to introduce the SemanticFrame > class is that it is somehow non-standard to say that a Lexical > Sense has semanticArguments. This will be strange for many people. > It will be much clearer if we say that a SemanticFrame has > semantic arguments, where the SemanticFrames simply stands proxy > for a certain ontological configuration in the ontology. > > So what I am proposing is to redefine the property semArg to have > SemanticFrame as domain, and making SemanticFrame a subclass of > Sense. In some sense a SemanticFrame is thus a special case of a > Sense that is a gestalt-like thing having semantic arguments. > > The model is increased by one class, true, that is really the only > drawback I see. But it makes the model conceptually clearer and > more accessible I believe. The advantage is that this extension is > compatible with previous versions. If people stick to the previous > modelling, the only consequence is that the LexicalSenses the have > been using so far will be inferred to be SemanticFrames. This does > not intefere with anyhting they have done and produces the desired > inference. > > Regards, > > Philipp. > > > > Am 10.07.14 11:37, schrieb Armando Stellato: >> >> Dear all, >> >> my (really poor) two cents: >> >> I agree mostly with John, except that, well, yes, I wouldn’t be >> so close wrt introducing frames ion general. But I suspect this >> is again a matter of principle: either we want to *only* have a >> model which coherently depicts things in a given way, or we may >> **also** want to represent existing resources according to it. >> One of the things in the limbo between the two approaches has >> always been the representation of existing lexical resources. >> This is, by definition, not in the scope of OntoLex, though, in >> the absence of existing RDF models for lexical resources, >> inevitably (IMHO) it should be addressed. >> >> So, to me it wouldn’t be bad to have a frame resources module, >> and I see a SemanticFrame in there. Again, my preference goes to >> have the possibility of seeing existing resources not depicted by >> their own ontology (e.g. FrameNet ontology), but rather seen >> under a larger umbrella. >> >> However, I don’t see any kind of inclusion (in a sense or the >> other) with LexicalSense, and I better see it as a separate object. >> >> Cheers, >> >> Armando >> >> *From:*johnmccrae@gmail.com <mailto:johnmccrae@gmail.com> >> [mailto:johnmccrae@gmail.com] *On Behalf Of *John P. McCrae >> *Sent:* Thursday, July 10, 2014 11:12 AM >> *To:* Philipp Cimiano; Philipp Cimiano >> *Cc:* public-ontolex@w3.org <mailto:public-ontolex@w3.org>; >> public-ontolex@w3.org <mailto:public-ontolex@w3.org> >> *Subject:* Re: synsem module >> >> Hi, >> >> On Wed, Jul 9, 2014 at 12:32 PM, Philipp Cimiano >> <cimiano@cit-ec.uni-bielefeld.de >> <mailto:cimiano@cit-ec.uni-bielefeld.de>> wrote: >> >> Dear all, >> >> I am working through the synsem module, see my updates on >> the GIT repository. >> >> I do not have major changes of this module other than the >> following two: >> >> 1) I have changed a number of definitions to make them >> clearer, please check and let me know if the definitions are >> fine. >> >> 2) For the sake of symmetry, I propose to add a class >> "SemanticFrame" as a counterpart to Frame, which represents a >> syntactic frame, essentially capturing the valence or subcat >> behaviour of a given lexical entry. This SemanticFrame would >> essentially be a subclass of LexicalSense, and would leave >> the other parts of the model essentially untouched. I feel >> that having this symmetry (syntactic and semantic side) makes >> the model more elegant and clearer. Some people will be >> looking for something like this. Essentially, a SemanticFrame >> would represent a gestalt-like conceptual construction that >> represents the meaning of a lexical entry. >> >> I have chosen the following definition for the >> "SemanticFrame" class: A Semantic Frame is a coherent >> structure of related concepts that are related such that >> without knowledge of all of them, one does not have complete >> knowledge of any one; they are in that sense types of >> gestalt. The coherent structure is represented by one or more >> predicates from a given ontology. >> >> I'm not sure what this brings us, it adds an extra class (which >> inevitably increases complexity and confusion) for no technical >> advantage. That is do we really have a concrete example where it >> would be good to use a SemanticFrame instead of a LexicalSense? >> >> Also, I am not sure that the axiomatization of SemanticFrame as a >> subclass of LexicalSense makes sense... in particular is it not >> the case that every LexicalSense is a SemanticFrame as it refers >> to a concept in the ontology and is thus simply mapped to the >> argument structure of the ontological predicate, thus every >> lexical sense necessarily is associated with a semantic frame. If >> we agree that SemanticFrame ⊒ LexicalSense, we should then ask is >> there is a semantic frame that is not a lexical sense? Firstly, >> from the point of view of OntoLex *all semantic is in the >> ontology*, therefore a semantic frame must also refer to the >> ontology, thus we need only ask if there is such a thing as a >> /non-lexicalized/ semantic frame? The conclusion that was reached >> in Monnet was that there was no such thing, or at least such a >> thing is not relevant is not to OntoLex (as we only wish to >> describe how ontologies are lexicalized), thus we could say that >> LexicalSense ≡ SemanticFrame and eliminate the unnecessary >> synonym from the model. >> >> From a strategic standpoint, I think that we should avoid adding >> the semantic frame in because "people will be looking for >> something like this". The fact that people will look for this >> means that if they find something with a name like this that >> doesn't actually work like they expect then they are guaranteed >> to misuse it! Instead, if they find a clear documentation of why >> such an object does not exist (i.e, "semantics is in the >> ontology") then that will help them far more than introducing a >> confusing subclass. >> >> The definition as it stands currently is also weak for similar >> reasons... if a semantic frame is a "structure represented by one >> or more predicates from an ontology", why is it in the lexicon >> not entirely in the ontology?? >> >> Regards, >> >> John >> >> >> Please check the ontology, the examples etc. and help me to >> debug the ontology, description and examples. >> >> Best regards, >> >> Philipp. >> >> >> -- >> >> >> >> Prof. Dr. Philipp Cimiano >> >> >> >> Phone:+49 521 106 12249 <tel:%2B49%20521%20106%2012249> >> >> Fax:+49 521 106 12412 <tel:%2B49%20521%20106%2012412> >> >> Mail:cimiano@cit-ec.uni-bielefeld.de <mailto:cimiano@cit-ec.uni-bielefeld.de> >> >> >> >> Forschungsbau Intelligente Systeme (FBIIS) >> >> Raum 2.307 >> >> Universität Bielefeld >> >> Inspiration 1 >> >> 33619 Bielefeld >> > > -- > -- > Prof. Dr. Philipp Cimiano > AG Semantic Computing > Exzellenzcluster für Cognitive Interaction Technology (CITEC) > Universität Bielefeld > > Tel:+49 521 106 12249 <tel:%2B49%20521%20106%2012249> > Fax:+49 521 106 6560 <tel:%2B49%20521%20106%206560> > Mail:cimiano@cit-ec.uni-bielefeld.de <mailto:cimiano@cit-ec.uni-bielefeld.de> > > Office CITEC-2.307 > Universitätsstr. 21-25 > 33615 Bielefeld, NRW > Germany > > -- -- 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
Received on Friday, 18 July 2014 12:02:28 UTC