W3C home > Mailing lists > Public > public-media-annotation@w3.org > June 2010

RE: [mawg] [rdfs taskforce] action-249: RDFS version of ma-ont availble

From: Chris Poppe <Chris.Poppe@ugent.be>
Date: Fri, 11 Jun 2010 15:40:41 +0200
To: "'Evain, Jean-Pierre'" <evain@ebu.ch>, "'Bailer, Werner'" <werner.bailer@joanneum.at>, 'Tobias Bürger' <tobias.buerger@salzburgresearch.at>, <johns@postech.ac.kr>
Cc: <public-media-annotation@w3.org>
Message-ID: <02a001cb096b$afdda400$0f98ec00$@Poppe@ugent.be>
Dear all,

For me the choice between subproperties or properties modeled as class is just an ontology engineering issue. Whatever the choice, the same information will be available. 
However, I do think that the information defined in the ontology document should be represented in the rdfs/owl ontology. 

So if we opt for subproperties, it would be good to add such subproperties to the ontology (as the example you give). This will make it clear to others how different types of identifiers (and other properties) can be included.

Using this as a guideline, the role for Creator could also be a subproperty of hasContributor   (for instance for the role director we have an objectproperty hasDirector, which is a subproperty of hasContributor). 

For the framesize I agree about the blank node, however, we should include the units of the framesize somehow, since it is also in the ontology document.

Kind regards,
Chris



-----Original Message-----
From: Evain, Jean-Pierre [mailto:evain@ebu.ch] 
Sent: vrijdag 11 juni 2010 15:21
To: Evain, Jean-Pierre; 'Bailer, Werner'; Chris.Poppe@UGent.be; 'Tobias Bürger'; johns@postech.ac.kr
Cc: public-media-annotation@w3.org
Subject: RE : [mawg] [rdfs taskforce] action-249: RDFS version of ma-ont availble

Dear all,

I have a few more minutes.

As for the identifier and type, you could have an 'identifier' data property of anyURI and subproperties for the type of identifiers. Eg. ISAN could be a subdataproperty of identifier ( anyURI type again).

Coming back to frameSize-> doing a class frameSize with two dataproperties hieght and width was my first approach when I started RDF. -> this takes you into defining a blank node. Generating individuals will just ignore this dummy class. It is clearly  not recommended and this adds extra complications in writing queries. Even from a modle perspective, it is not clean.

About persons and organisations being subclass or contributor, etc.  Actually the MAWG only models persons and organisations to instantiate contributors, creators, etc. there is nothing like an individual person or organisation. However, I can't remember if I have them at both level but this is not a big issue.

Person doesn't need to be under 'thing'. I thought I had removed most of them.

Example of SKOS: any skos concept is identifier by its ID coinciding with the URI+ termId of the e.g. mpeg or EBU lists. this gives you access to the properties like preferredLabel, etc.

Regards,

jean-pierre
________________________________
De : public-media-annotation-request@w3.org [public-media-annotation-request@w3.org] de la part de Evain, Jean-Pierre [evain@ebu.ch] Date d'envoi : vendredi, 11. juin 2010 14:54   : 'Bailer, Werner'; Chris.Poppe@UGent.be; 'Tobias B rger'; johns@postech.ac.kr Cc : public-media-annotation@w3.org Objet : RE: [mawg] [rdfs taskforce] action-249: RDFS version of ma-ont availble

Dear all, Chris,

I have been reviewing the RDFS file and not what is on line. There may be syntax issues that makes that my RDF parser sees some dataproperties as object properties as well as datatypes appearing as 'classes'.

I totally disagree with having properties like identifier and title becoming classes. For me, this is heresy ;-) Even worse for frameSize. Absolute nonsense! -> just keep it flat with two values.

You are shooting yourself in the foot and drifting away from the class model of the ontology-> really really weird! (not to say insane !!)

Genre and keyword can be either a data property or an object property is using SKOS concept classes.

Whether mediaFragment is a class will depend whether it can be identified as a resource. My feeling is that using the ID of the resource extended by a pointer will serve as a valid ID making mediaFragments eligible as classes, hence subclass of resource.

'main' as subproperty of title, I agree we can do it now. That's what I suggested as being the working rules for translating 'types' (of titles, descriptions) into RDF

Contributor, creators: role is either a data property (string) or object property (SKOS concept) of the class. I think I have implemented it in my revised version

Etc. and many others. BUT PLEASE DON?T MAKE CLASS OUT OF PROPERTIES

Regards,

Jean-Pierre




From: Bailer, Werner [mailto:werner.bailer@joanneum.at]
Sent: vendredi, 11. juin 2010 14:38
To: Chris.Poppe@UGent.be; Evain, Jean-Pierre; 'Tobias B rger'; johns@postech.ac.kr
Cc: public-media-annotation@w3.org
Subject: AW: [mawg] [rdfs taskforce] action-249: RDFS version of ma-ont availble

Dear Tobias, Chris, Jean-Pierre, all,

I generally agree to Chris  comments, and I d like to add the following:

1. as Chris has mentioned, there are several properties which have type and value, and could be modelled as classes. I think this applies to the following properties: Identifier, title, createDate, policy, copyright, targetAudience, rating I m not sure that using subproperties can work for all of them, e.g. how would you deal with the 4 attributes of rating, and would you want define a subproperty for each identifier of copyright or policy?

frameSize (width, height, unit) is now represented as frameWidth, frameHeight   I m not sure why we should model this different than other complex properties, so I d suggest to model it as a class with its 3 properties

keyword and genre are defined to be a string OR a URI, the current definition with skos:Concept is more strict as it excludes the string option.

2. Person/Organisation: I agree with Jean-Pierre s and Chris  comments, and part of the issues coming fromm mixing two dimensions of specialisation, namely into Person/Organisation and Creator/Contributor/Publisher. According to our ontology document we do not discriminate between persons and organisations, so an Agent specialised into Creator/Contributor/Publisher would be sufficient to represent what we have defined in our ontology document.

3. I agree that a MediaFragment should be a subclass of MediaResource. I m also not sure whether we need Track as a separate class, as they are MediaFragments, but if we have a separate class, it should be a subclass of MediaFragment.

Best regards,
Werner

Von: public-media-annotation-request@w3.org [mailto:public-media-annotation-request@w3.org] Im Auftrag von Chris Poppe
Gesendet: Freitag, 11. Juni 2010 08:54
An: 'Evain, Jean-Pierre'; 'Tobias B rger'; johns@postech.ac.kr
Cc: public-media-annotation@w3.org
Betreff: RE: [mawg] [rdfs taskforce] action-249: RDFS version of ma-ont availble

In the RDFS ontology, I assume that Audio and Image, should also be subclassed from MediaResource.

General comments, starting from the Ontology document [1]:

          Should the ontology more closely represent the  ontology  described in the Ontology document?

          The ma:identifier property holds both the actual id and a category of the identifier (see [1]).
 So would it make sense to make identifier an ObjectProperty pointing to an Identifier  Class, which has two dataproperties ( id, range: URI and type, range: String)?

          The same holds for title and all other  subtype properties . I read somewhere that you want to use subproperties for types of titles. E.g., make mainTitle a subproperty of the title dataproperty?
It would be good to add such a subproperty already to the ontology.

          I find it strange that Persons and organizations are subclasses of Creators, Publishers, and Contributors. IMO this makes no sense, because this states that all persons are Creators, Publishers and Contributors  I would prefer to include Agent taking Persons, Organizations as subclass (like in FOAF). (Is there a reason for not reusing FOAF classes?).
Additionally, I would make Contributor a subclass of Agent. So if we have a Person Instance Person_0 and add a  MediaResource_0 hasContributor Person_0  it can be deduced that Person_0 is both a person and contributor (the same is possible for organizations).

          How is the role defined for contributors/creators?

          When I open the ontology in Prot g , I see the Person classes at different places, is it necessary to explicitly state that Person is a subclass of Thing , or was this generated automatically?

          How will the different types of dates be defined?

          Could you give an example of instance data for the SKOS:Concept used for genre/keyword/rating/targetaudience?

          For the collections, inverse properties would be usefull

          For the fragments, how about making MediaFragment a subclass from MediaResource?

          Add a label dataproperty, range:string to the NamedMediaFragment Class

          Add the technical properties to the correct subclass of MediaResource (e.g., frameWidth and frameHeight for Image and Video, samplingRate for Audio)

          I think it would be nice to add annotations to the different classes, properties, and so on.

          Should these tracks be included as separate classes? Are these also MediaResources, or even MediaFragments? The ontologydoc does not include information on the tracks themselves, only on the number ?
Some comments on Jean-Pierre s comments I added inline below ([CP])


Kind regards,
Chris



[1]  http://www.w3.org/TR/2010/WD-mediaont-10-20100608/






From: public-media-annotation-request@w3.org [mailto:public-media-annotation-request@w3.org] On Behalf Of Evain, Jean-Pierre
Sent: maandag 7 juni 2010 11:32
To: 'Tobias B rger'; johns@postech.ac.kr
Cc: public-media-annotation@w3.org
Subject: RE: [mawg] [rdfs taskforce] action-249: RDFS version of ma-ont availble

Hello Tobias,

A few comments:


-          When looking at the ontology in Prot g , datatype like integer appear as classes  That is weird.

[CP] I don t have this problem, Which version of Prot g  are you using?


-          This is why you have what I believe is a wrong use of 'object properties'.  For example, bitrate should be a dataproperty  of media resource instantiated by an integer (not a float ;-)

[CP] (In the version that I see online, bitrate is  a dataproperty) In fact, the ontology document states that bitrate is a Float, maybe we should also rename this property to averageBitrate as in the ontology document?


-          Language is not a class but an object property linking to a SKOS concept or a data property linking to a string -> the same for e.g. compression, format, genre, keyword, rating and targetAudience


[CP] In the version I have, language is a dataproperty, so I am beginning to think that I am reviewing a revised version?


-          Person should also be a subclass of Actor and probably organisation could also appear where person is.

-          The flattest the better: I would avoid nesting too much things like audio/video, tracks and then audio/video tracks

-          partOfCollection could be replaced by a more generic isMemberOf

-          relation would sound better like isRelatedTo and all Dublin Core relation could be subproperties

-          For the datatype of samplingRate, we often use rationals, which don't exist in RDF -> float?

-          Some properties are functional

-          Some properties are inverse, which I believe needs to be completed by proper object properties

-          Time to think of appropriate inverse properties for more inference

I have put a revised version here: http://www.ebu.ch/metadata/ontologies/W3C_MAWG/ma-ont-rev.rdfs.xml

To be discussed ?

Regards,

Jean-Pierre

From: public-media-annotation-request@w3.org [mailto:public-media-annotation-request@w3.org] On Behalf Of Tobias B rger
Sent: jeudi, 3. juin 2010 08:03
To: johns@postech.ac.kr
Cc: public-media-annotation@w3.org
Subject: Re: [mawg] [rdfs taskforce] action-249: RDFS version of ma-ont availble

Hi John,

starting next week is fine - every review is more than welcome!

Thank you in advance!

Best regards,

Tobias

Am 03.06.2010 03:08, schrieb Strassner John Charles:
Hallo Tobias,

I am happy to provide a review of both the RDFS and OWL documents; however, I cannot start the review until next week. I hope that is OK.

regards,
John
--- Original Message ---
>From : "Tobias B rger"<tobias.buerger@salzburgresearch.at><mailto:tobias.buerger@salzburgresearch.at>
To : "public-media-annotation@w3.org"<mailto:public-media-annotation@w3.org><public-media-annotation@w3.org><mailto:public-media-annotation@w3.org>
Date : 2010/06/02 Wednesday PM 5:51:18
Subject : [mawg] [rdfs taskforce] action-249: RDFS version of ma-ont availble

Dear all,

I have now also uploaded the RDFS version of the first implementation of ma-ont.

You can find it here: http://www.salzburgresearch.at/~tbuerger/ma-ont.rdfs
Browsable version:
http://liris.cnrs.fr/~pchampin/wsgi/t4r/?tu=ontoview.html&r=http%3A//www.w3.org/ns/ma-ont&gu=http%3A//www.salzburgresearch.at/~tbuerger/ma-ont.rdfs

I am still looking for a volunteer to provide a review and to suggest changes, extensions, etc. for both RDFS and OWL versions of ma-ont. Thanks.

Best regards,

Tobias

--
================================================================
Dr. Tobias B rger Knowledge and Media Technologies Group Salzburg Research FON +43.662.2288-415 Forschungsgesellschaft FAX +43.662.2288-222 Jakob-Haringer-Stra e 5/III tobias.buerger@salzburgresearch.at<mailto:tobias.buerger@salzburgresearch.at>
A-5020 Salzburg | AUSTRIA http://www.salzburgresearch.at




--

================================================================

Dr. Tobias B rger         Knowledge and Media Technologies Group

Salzburg Research                           FON +43.662.2288-415

Forschungsgesellschaft                      FAX +43.662.2288-222

Jakob-Haringer-Stra e 5/III   tobias.buerger@salzburgresearch.at<mailto: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 **************************************************

-----------------------------------------
**************************************************
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
**************************************************
Received on Friday, 11 June 2010 13:41:21 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Friday, 11 June 2010 13:41:22 GMT