RE : RE : ma-ont RDF latest version

We can probably force a rating provider  to provide only one rating value although if you wnat the same provider to provide different rating value per fragments, this is going to be a limitation and I don't think we wnat to define a rating provider per rating value. This is an implementaion issue. And we know RDF is not very good with cardinalities.

At first sight the two examples you are providing are valid... Version 1 isn't optimal and is also an implementation issue. It is a matter of database management or it shouldn't be forgotten that elements of description can come from different sources that give different IDs to their e.g. respective creators (in their database). The beauty of all this is that a query will always bring you back to YouTube. Isn't it?

JP






________________________________________
De : Höffernig, Martin [Martin.Hoeffernig@joanneum.at]
Date d'envoi : mardi, 2. novembre 2010 13:09
À : Bailer, Werner; Pierre-Antoine Champin; Chris Poppe
Cc : Evain, Jean-Pierre; Tobias Bürger; public-media-annotation@w3.org
Objet : AW: RE : ma-ont RDF latest version

Chris, thank you for providing an example!
I agree that the model is consistent. However, the fact that a RatingProvider can only give one rating is not formalized in the ontology spec yet. A way to achieve this would be to add cardinality restrictions on the concerning properties (ratingValue, ratingMin, ratingMax) in the class defintion of RatingProvider. (Of course cardinaltiy restrictions are OWL constructs and not possible using RDFS only.)

Anyway, if I understand the contributorIs property between a contributor and an agent rightly, than there a 2 different ways for example to describe that a given organisation is both the creator and the publisher of a movie:

Version1:
:YouTube a ma:Organization .

:YouTube1 a ma:Creator ;
        ma:creatorIs :YouTube .

:YouTube2 a ma:Publisher .
      ma:publisherIs YouTube .

:Movie a ma:MediaResource ;_
      ma:hasCreator :YouTube1 ;
      ma:hasPublisher :YouTube2 .

Version2:
:YouTube a ma:Organization .

:YouTube1 a ma:Creator ;
          a ma:Publisher ;
        ma:creatorIs :YouTube ;
        ma:publisherIs :YouTube .

:Movie a ma:MediaResource ;_
      ma:hasCreator :YouTube1 ;
      ma:hasPublisher :YouTube1 .

Please correct me if I am wrong with my assumptions.

Best,
Martin


-----Ursprüngliche Nachricht-----
Von: Bailer, Werner
Gesendet: Dienstag, 02. November 2010 11:46
An: Pierre-Antoine Champin; Chris Poppe
Cc: Evain, Jean-Pierre; Höffernig, Martin; Tobias Bürger; public-media-annotation@w3.org
Betreff: AW: RE : ma-ont RDF latest version

OK, I think Martin and I had the same misunderstanding.

In that case I think the model is ok, but it is misleading to call the class RatingProvider, which IMO would be the organisation, here it is rather a rating.

Best regards,
Werner

________________________________________
Von: Pierre-Antoine Champin [pierre-antoine.champin@liris.cnrs.fr]
Gesendet: Dienstag, 02. November 2010 11:07
An: Chris Poppe
Cc: Evain, Jean-Pierre; Bailer, Werner; Höffernig, Martin; Tobias Bürger; public-media-annotation@w3.org
Betreff: Re: RE : ma-ont RDF latest version

Oh, my bad indeed.
So I *was* mistaken by the class labels.

Ok on the general principle then.
I also agree with JP that RatingProvided (or whatever it is renamed to)
should not be a subclass of Contributor.

And this amounts to making Rating a class, IMHO.

   pa

On 11/02/2010 11:03 AM, Chris Poppe wrote:
> Dear all,
>
> as I understood a RatingProvider can only give one rating. It is
> connected to an Agent (Person or Organization) through the
> ratingProviderIs property.
> So something like this (?):
>
> :lmdb a ma:Organization ;
>
> :lmdb1 a ma:RatingProvider ;
>            ma:ratingMin 0 ;
>            ma:ratingMax 5;
>            ma:ratingProviderIs lmdb .
>
> :lmdb2 a ma:RatingProvider ;
>            ma:ratingMin 0 ;
>            ma:ratingMax 5 ;
>            ma:ratingProviderIs lmdb.
>
>      :movie1 ma:hasBeenRatedBy :lmdb1 .
>      :lmdb1 ma:ratingValue 3.
>
>      :movie2 ma:hasBeenRatedBy :lmdb2 .
>      :lmdb2 ma:ratingValue 5.
>
>
> Kind regards,
> Chris
>
> Quoting "Pierre-Antoine Champin"<pierre-antoine.champin@liris.cnrs.fr>:
>
>> On 11/01/2010 05:43 PM, Evain, Jean-Pierre wrote:
>>> Then I guess the easiest way is to also allow a property linking a
>>> rating provider to a fragment, which our new model allows.
>>
>> I think I agree with Marting and Werner that something there is a
>> problem in the current ontology (and I see it both in
>> TargetAudienceAuthory and RatingProvider).
>>
>> Imagine that I want to state that LinkedMDB rates movie1 3/5 and
>> movie2 5/5 . How would you state that in RDF?
>> As I understand the ontology, this would be
>>
>>    :lmdb a ma:RatingProvider ;
>>          ma:ratingMin 0 ;
>>          ma:ratingMax 5 .
>>
>>    :movie1 ma:hasBeenRatedBy :lmdb .
>>    :lmdb ma:ratingValue 3.
>>
>>    :movie2 ma:hasBeenRatedBy :lmdb .
>>    :lmdb ma:ratingValue 5.
>>
>> which is obviously broken, as the four last triples can be rewritten
>> like that:
>>
>>    :movie1 ma:hasBeenRatedBy :lmdb .
>>    :movie2 ma:hasBeenRatedBy :lmdb .
>>    :lmdb ma:ratingValue 3, 5.
>>
>> Again, the same problem raises with TargetAudienceAuthority.
>>
>> So either I'm mislead by the labels of the ontologies (in which case
>> I suggest they are renamed) or the ontology is broken...
>> I would prefer to write something like
>>
>>    :movie1 ma:hasRating [
>>      ma:ratingValue 3 ;
>>      ma:ratingMin 0 ;
>>      ma:ratingMax 5 ;
>>      ma:hasRatingAuthority :lmdb
>>    ]
>>
>> which is much closer to the Json specified by the API -- and yes, it
>> amounts to define a class for ratings. But frankly, I don't see any
>> other way to convey the same information as the API...
>>
>>    pa
>>
>>>
>>> Don't you think so?
>>>
>>> Regards, JP
>>>
>>> ________________________________________ De : Bailer, Werner
>>> [werner.bailer@joanneum.at] Date d'envoi : lundi, 1. novembre 2010
>>> 17:27 À : Evain, Jean-Pierre; Höffernig, Martin; 'Chris Poppe' Cc :
>>> Tobias Bürger; public-media-annotation@w3.org Objet : RE: ma-ont RDF
>>> latest version
>>>
>>> Dear Jean-Pierre,
>>>
>>> I agree that describing rating providers is out of scope of MAWG. The
>>> motivation behind Martin's comment was the following scenario: Assume
>>> you have one RDF graph that containing the description of media
>>> resource and its fragments (resources themselves). Different of the
>>> fragments got different ratings from the same provider - how could
>>> you describe that? hasRated would always point to the same
>>> RatingProvider instance.
>>>
>>> Best regards, Werner
>>>
>>>> -----Original Message----- From: Evain, Jean-Pierre
>>>> [mailto:evain@ebu.ch] Sent: Samstag, 30. Oktober 2010 12:29 To:
>>>> Evain, Jean-Pierre; Höffernig, Martin; 'Chris Poppe' Cc: Tobias
>>>> Bürger; Bailer, Werner; public-media-annotation@w3.org Subject: RE:
>>>> ma-ont RDF latest version
>>>>
>>>> Dear Martin,
>>>>
>>>> As promised, let's continue the discussion about the rating value.
>>>>
>>>> 1. If I understand well, the intention is likely to be able to
>>>> find other resources that the rating provider might have reviewed
>>>> because e.g. a user finds his rating accurate and expect finding
>>>> other content of interest based on the ranking of the rating
>>>> provider. Right? If yes, then the current ontology allows making
>>>> queries on all resources rated by the rating provider even possibly
>>>> adding a filter on certain rating values.
>>>>
>>>> 2. If the intention of your comment is to develop an ontology for
>>>> the description of rating providers listing all their ratings, this
>>>> is not (at least directly - I believe) within the scope of the
>>>> MAWG.
>>>>
>>>> 3. There is a fundamental modelling issue with your proposal to
>>>> have a rating value to which would be associated properties by the
>>>> rating provider definition.  This would require a rating value to
>>>> be a class and it is not advisable (again - I believe) to make a
>>>> class of what is fundamentally a property. A question to help
>>>> sorting this out: would you have a database in which you would
>>>> order the information per rating value (each of them would then
>>>> have an identifier, which could be used to relate to them as
>>>> classes)? - of course I have my own opinion but would like to hear
>>>> yours ;-)
>>>>
>>>> In conclusion, my gut feeling is that the current representation
>>>> is accurate from a modelling perspective. Do you really believe
>>>> that it is not considering my explanations above?
>>>>
>>>> Best regards,
>>>>
>>>> Jean-Pierre ________________________________________ De :
>>>> Höffernig, Martin [Martin.Hoeffernig@joanneum.at] Date d'envoi :
>>>> jeudi, 28. octobre 2010 14:27 À : Evain, Jean-Pierre; 'Chris
>>>> Poppe' Cc : Tobias Bürger; Bailer, Werner;
>>>> public-media-annotation@w3.org Objet : AW: ma-ont RDF latest
>>>> version
>>>>
>>>> Dear Jean-Pierre, Chris and all,
>>>>
>>>> just a few comments regarding the current ontology spec:
>>>>
>>>> Since TargetAudienceAuthority is no longer a sub class of
>>>> Contributor the object property targetAudienceAuthorityIs shouldn't
>>>> be a sub property of contributorIs as well. Leaving this sub
>>>> property relation would infer that the domain of property
>>>> targetAudienceAuthorityIs is Contributor, since Contributor is
>>>> domain of property contributorIs.
>>>>
>>>> As Chris wrote there is no direct connection between a
>>>> MediaResource and the value of TargetAudienceAuthority, the same
>>>> problem/pattern exists between a MediaResource and a given rating
>>>> value. Since a rating value is directly related to a RatingProvider
>>>> (via data property ratingValue), it is not expressible that the
>>>> same RatingProvider rates different MediaResources (or
>>>> MediaFragments) including different rating values. I propose to
>>>> directly connect a rating value to a MediaResource and annotate
>>>> this rating value with additional information (RatingProvider,
>>>> etc.). A new class RatingValue would be necessary for this
>>>> purpose.
>>>>
>>>> Best, Martin
>>>>
>>>>
>>>> -----Ursprüngliche Nachricht----- Von:
>>>> public-media-annotation-request@w3.org [mailto:public-media-
>>>> annotation-request@w3.org] Im Auftrag von Evain, Jean-Pierre
>>>> Gesendet: Montag, 25. Oktober 2010 10:11 An: 'Chris Poppe' Cc:
>>>> Tobias Bürger; Bailer, Werner; Davy Van Deursen; public-media-
>>>> annotation@w3.org Betreff: RE: ma-ont RDF latest version
>>>>
>>>> Chris,
>>>>
>>>> As below and attached... Hope this answers the questions and
>>>> needs.
>>>>
>>>> Regards,
>>>>
>>>> Jean-Pierre
>>>>
>>>>
>>>>
>>>> -----Original Message----- From: Chris Poppe
>>>> [mailto:Chris.Poppe@UGent.be] Sent: dimanche, 24. octobre 2010
>>>> 21:10 To: Evain, Jean-Pierre Cc: Tobias Bürger; Bailer, Werner;
>>>> Davy Van Deursen; public-media- annotation@w3.org Subject: RE:
>>>> ma-ont RDF latest version
>>>>
>>>> Dear all,
>>>>
>>>> congrats with the excellent work, seems like we have a real
>>>> ontology instead of a property list now :).
>>>>
>>>> Some remarks: In the ontology specification the location property
>>>> is defined as the location where a resource is created, developed,
>>>> recorded, or otherwise authored. Currently the ontology scheme only
>>>> has a depictedLocation. So maybe a ObjectProperty createLocation
>>>> could be added? I guess the depicted location is not in the
>>>> ontology specification since it could be described using the
>>>> "description" property?
>>>>
>>>> JPE: Although a location could be described in 'description', it is
>>>> the function of location to do this also as a linked data hook (and
>>>> I would says it would make more sense to insist on what is shown
>>>> that where it was developed?!?!). But how? It seems the semantics
>>>> gives a list of properties. We could have a general property like
>>>> 'hasRelatedLocation' (which would be as vague as the way it is
>>>> currently defined) as a placeholder to Develop a series of
>>>> subproperties: depicted, created, developed, etc.
>>>>
>>>>
>>>> Could isImageRelatedTo be made a subproperty of isRelatedTo?
>>>>
>>>> JPE: No, not the same domain. But hasRelatedImage could be.
>>>>
>>>>
>>>> Could the link between the Data Property relation and the Object
>>>> Peropty isRelatedTo be formalized somehow? E.g., if a MediaResource
>>>> is used as the range of a isRelatedTo objectproperty, it implies
>>>> that a relation data property should exist with the URI of that
>>>> MediaResource?
>>>>
>>>> JPE: Good point.  Actually, if the 'relation' is e.g. 'source'
>>>> (the source from which the mediaresource is derived) , then 'source
>>>> should be a subproperty of isRelatedTo. I have used this as an
>>>> example and removed the 'relation' dataproperty.
>>>>
>>>>
>>>> There is no direct connection between a MediaResource and the
>>>> TargetAudienceAuthority. Do I interpret it correct that to express
>>>> that an organization has given a classification "adult" to a
>>>> mediaResource, we express this as: a_MediaResource hasContributor
>>>> a_TargetAudienceAuthority; a_TargetAudienceAuthority
>>>> targetAudienceAuthorityIs a_Organization; a_TargetAudienceAuthority
>>>> targetAudience "Adult";
>>>>
>>>> JPE: 1/ Yes, a few relations to some contributors were missing
>>>> inc. targetAudienceAuthority 2/ Yes again, the only way to express
>>>> different targetAudience e.g. using different target audience
>>>> schemes was to use the trick of linking the property to the
>>>> authority. BTW, I would suggest the authority is no longer a
>>>> subclass of contributor ->   agree?
>>>>
>>>>
>>>> I think there is something missing to state that a MediaFragment is
>>>> a fragment of a specific MediaResource (maybe isFragmentOf and
>>>> hasFragment Object properties).
>>>>
>>>> JPE: Yes
>>>>
>>>> Maybe it's better to remove the namedFragmentUri data property and
>>>> create a fragmentName data property (like the fragmentRole
>>>> property). This way a namedFragment is a MediaFragment with a
>>>> fragmentName and the URI can be retrieved through fragmentUri.
>>>>
>>>> JPE: That's what I suggested in a previous mail. Actually I
>>>> corrected the existing mediaFragmentName data property into
>>>> fragmentName. I have now kept only locator and renamed fragmentUri
>>>> into fragmentLocator (or we change locator in mediaResourceUri
>>>> :-).
>>>>
>>>> Kind regards, Chris
>>>>
>>>> Quoting "Evain, Jean-Pierre"<evain@ebu.ch>:
>>>>
>>>>> Dear Thierry,
>>>>>
>>>>> Almost a week without additional comment. I would therefore
>>>>> suggest that this becomes the new version of our RDF ontology.
>>>>>
>>>>> Thanks in advance for uploading it and replace the current
>>>>> published
>>>> version.
>>>>>
>>>>> Best regards,
>>>>>
>>>>> Jean-Pierre
>>>>>
>>>>>
>>>>>
>>>>> -----Original Message----- From: Evain, Jean-Pierre Sent:
>>>>> vendredi, 15. octobre 2010 06:19 To: Evain, Jean-Pierre; Tobias
>>>>> Bürger; Bailer, Werner Cc: Davy Van Deursen;
>>>>> public-media-annotation@w3.org Subject: RE : ma-ont RDF latest
>>>>> version
>>>>>
>>>>> Dear all,
>>>>>
>>>>> this is the new version with MediaFragment as a subclass of
>>>>> MediaResource, validated as OWL-DL.
>>>>>
>>>>> Please check and feedback.
>>>>>
>>>>> Best regards,
>>>>>
>>>>> Jean-Pierre
>>>>>
>>>>>
>>>>> ________________________________________ De : Evain, Jean-Pierre
>>>>> Date d'envoi : vendredi, 15. octobre 2010 02:09 À : Tobias
>>>>> Bürger; Bailer, Werner Cc : Davy Van Deursen;
>>>>> public-media-annotation@w3.org Objet : RE : ma-ont RDF latest
>>>>> version
>>>>>
>>>>> Thanks Tobias, all,
>>>>>
>>>>> There seem to be concensus. I'll work on a new version.
>>>>>
>>>>> I was thinking about namedFragment. Although the MFWG makes this
>>>>> disctinction, I wonder if we need to in MAWG as we would have a
>>>>> property 'name' that be be documented or not.  Then the URI
>>>> attributed
>>>>> to the fragment would use an MFWG format or another,
>>>>> accordingly.
>>>>>
>>>>> I hope I'll find 5 minutes to do this today during my various
>>>> meetings.
>>>>>
>>>>> Regards,
>>>>>
>>>>> Jean-Pierre
>>>>>
>>>>>
>>>>> ________________________________________ De : Tobias Bürger
>>>>> [tobias.buerger@salzburgresearch.at] Date d'envoi : jeudi, 14.
>>>>> octobre 2010 18:20 À : Bailer, Werner Cc : Evain, Jean-Pierre;
>>>>> Davy Van Deursen; public-media-annotation@w3.org Objet : Re:
>>>>> ma-ont RDF latest version
>>>>>
>>>>> Dear all,
>>>>>
>>>>> given the definition of MF cited below, it makes sense to model
>>>>> MF
>>>> like that.
>>>>>
>>>>> Best,
>>>>>
>>>>> Tobias
>>>>>
>>>>> Am 14.10.2010 15:34, schrieb Bailer, Werner:
>>>>>> Dear Davy, Jean-Pierre, all,
>>>>>>
>>>>>> I agree with the proposal that a media fragment is a subclass
>>>>>> of media resource.
>>>>>>
>>>>>> Actually, this a clean way of modeling it, as we anyway
>>>>>> couldn't prevent someone from expressing that by using a MFURI
>>>>>> as the URI of
>>>> a
>>>>>> media resource.
>>>>>>
>>>>>> Best regards, Werner
>>>>>>
>>>>>>
>>>>>>> -----Original Message----- From:
>>>>>>> public-media-annotation-request@w3.org [mailto:public-media-
>>>>>>> annotation-request@w3.org] On Behalf Of Evain, Jean-Pierre
>>>>>>> Sent: Donnerstag, 14. Oktober 2010 15:25 To: Davy Van
>>>>>>> Deursen Cc: public-media-annotation@w3.org Subject: RE :
>>>>>>> ma-ont RDF latest version
>>>>>>>
>>>>>>> Hi Davy,
>>>>>>>
>>>>>>> Thank for summarsing the semantics, that will help me
>>>>>>> answering the question... (I hope :-)
>>>>>>>
>>>>>>> [[ Therefore, we should first look at the definition of a
>>>>>>> media resource [1] and I believe that a media fragment falls
>>>>>>> under that definition (if not, please clarify why not): " A
>>>>>>> media resource is any physical or logical Resource that can
>>>>>>> be identified using a Uniform Resource Identifier (URI), as
>>>>>>> defined by [RFC 3986]) , which has or is related to one or
>>>>>>> more media content types."  More specifically, a media
>>>>>>> fragment is a physical
>>>> resource,
>>>>>>> with a media content type (i.e., the same as its parent
>>>>>>> resource) and can be identified using a URI (i.e., a Media
>>>> Fragments
>>>>>>> URI).]]
>>>>>>>
>>>>>>> This is effectively the key question and I would inviote the
>>>>>>> whole MAWG to consider this question.
>>>>>>>
>>>>>>> My first intention would have been to have media fragment as
>>>>>>> a subclass of media resource composed of audio and video
>>>>>>> tracks. If
>>>> we
>>>>>>> all adopt and recognise more specifically that a fragment is
>>>>>>> a
>>>> media
>>>>>>> resource which is iodentified by a MFURI I am happy with this
>>>>>>> but the group needs to confirm what the mediaFragment is.
>>>>>>> Then we could name (namedFragment, itself a subclass of
>>>>>>> fragment) and keyword a fragment and give him a URI. That
>>>>>>> would be 'clean'.
>>>>>>>
>>>>>>> Then  if the question arises of whether a media fragment is
>>>>>>> a subclass of media resource, I would answer that any media
>>>>>>> resource is an atomic media fragment.
>>>>>>>
>>>>>>> In other words, I personally can agree with what you suggest
>>>>>>> but would like to hear from the group.
>>>>>>>
>>>>>>> Tobias and team, what do you think?
>>>>>>>
>>>>>>> Best regards,
>>>>>>>
>>>>>>> Jean-Pierre -----------------------------------------
>>>>>>> ************************************************** This email
>>>>>>> and any files transmitted with it are confidential and
>>>>>>> intended solely for the use of the individual or entity to
>>>>>>> whom
>>>> they
>>>>>>> are addressed. If you have received this email in error,
>>>>>>> please notify the system manager. This footnote also confirms
>>>>>>> that this email message has been swept by the mailgateway
>>>>>>> **************************************************
>>>>>>>
>>>>>>
>>>>>
>>>>> --
>>>>> ================================================================
>>>>> Dr. Tobias Bürger         Knowledge and Media Technologies Group
>>>>> Salzburg Research                           FON +43.662.2288-415
>>>>> Forschungsgesellschaft                      FAX +43.662.2288-222
>>>>> Jakob-Haringer-Straße 5/III   tobias.buerger@salzburgresearch.at
>>>>> A-5020 Salzburg | AUSTRIA         http://www.salzburgresearch.at
>>>>>
>>>>>
>>>>> -----------------------------------------
>>>>> ************************************************** This email and
>>>>> any files transmitted with it are confidential and intended
>>>>> solely for the use of the individual or entity to whom they are
>>>>> addressed. If you have received this email in error, please
>>>>> notify the system manager. This footnote also confirms that this
>>>>> email message has been swept by the mailgateway
>>>>> **************************************************
>>>>>
>>>>
>>>>
>>>>
>>>> -- Ghent University - Multimedia Lab Sint-Pietersnieuwstraat 41
>>>> B-9000 Ghent, Belgium
>>>>
>>>> tel: +32 9 264 89 17 fax: +32 9 264 35 94 e-mail:
>>>> Chris.Poppe@ugent.be
>>>>
>>>> URL: http://multimedialab.elis.ugent.be
>>>
>>
>>
>
>
>
> --
> Ghent University - Multimedia Lab
> Sint-Pietersnieuwstraat 41
> B-9000 Ghent, Belgium
>
> tel: +32 9 264 89 17
> fax: +32 9 264 35 94
> e-mail: Chris.Poppe@ugent.be
>
> URL: http://multimedialab.elis.ugent.be
>
>
>

Received on Tuesday, 2 November 2010 13:13:17 UTC