Re: my token about the "3 or more layer" structure for the ontology

Pierre-Antoine Champin さんは書きました:
> Felix Sasaki a écrit :
>   
>> Hi Pierre-Antoine again (btw. ,do you mind to forward this to the
>> mailing list?),
>>     
>
> absolutely not! I just missed the "reply all" button in my last reply :-(
> Sorry everybody... However, the last exchange with Felix in enclosed below.
>
>   
>> I think we don't disagree - we just stress different points.
>>     
>
> It is also my opinion :)
>   

Great :)

> And by the way, I am perfectly ok with the bottom-up approach, having
> the conceptual model emerge from use cases, which is indeed a good way
> of keeping it (the conceptual model) simple enough.
>
> Thinking about it, I realized that external metadata is indeed an
> important feature for me,

Could you give an example why? I.e., a practical "application scenario" 
In the sense of
http://lists.w3.org/Archives/Public/public-media-annotation/2008Nov/0100.html
(if we go for that structuring proposal)
Allowing for external metadata would be a use case IMO, based on an 
application scenario. The requirement would be to provide the external 
format.

>  hence my bias in favor of a specific format
> for our ontology.
>
> External metadata allows people to exchange it separately from the
> (potentially copyrighted) video. It is indeed a key feature of the
> Advene application (on which I am working), as well as popular social
> applications focused on video (like youtube).
>
> Wouldn't that deserve a use-case, by the way?
>   

Yes please, see above. We also have a "mini" issue about this already: 
http://www.w3.org/Bugs/Public/show_bug.cgi?id=6109

Felix

>   pa
>
>   
>> Pierre-Antoine Champin さんは書きました:
>>     
>>> Felix,
>>>
>>> Felix Sasaki a écrit :
>>>  
>>>       
>>>> Pierre-Antoine Champin さんは書きました:
>>>>    
>>>>         
>>>>> Hi Felix, thank you for your feedback.
>>>>>
>>>>> First, the term "data structure" was a bad choice. I should have
>>>>> written
>>>>> "conceptual model", which describes better what I am interested in. I
>>>>> think once we agree on a conceptual model, we can chose the best syntax
>>>>> to represent it -- if we want to...
>>>>>
>>>>> As a matter of fact, I took for granted that we would have to define
>>>>> our
>>>>> own format. But once again, the most important thing is the conceptual
>>>>> model.
>>>>>         
>>>>>           
>>>> I'm sorry, I have to disagree again. Take again a statement like
>>>> "Mapping for getDate" from
>>>> http://dev.w3.org/2008/video/mediaann/mediaont-api-1.0/mediaont-api-1.0.html#property-createDate
>>>>
>>>>
>>>> "exif:DateTimeDigitized (based on [MWG Guidelines Image]). This is taken
>>>> only into account if both xmp:createDate and exif:DateTimeOriginal are
>>>> missing."
>>>> I think this fulfils our task without a conceptual model: to provide
>>>> interoperability between heterogenous metadata formats for video.
>>>>     
>>>>         
>>> how do you (or the MWG Guidelines) end up with such a recommendation, if
>>> not by building (even if only mentally) a conceptual model of the
>>> metadata, beyond a simple attribute-value model?
>>>   
>>>       
>> You hit the point with "even mentally". I'm fine with a conceptual
>> model. However, I am worried that we spend too much time thinking about
>> a conceptual, without working on our main task of describing
>> interoperability between properties of existing formats. IMO a
>> conceptual model, more ore less explicit, will come naturally along if
>> we concentrate on that task.
>>
>> If we concentrate on the model we might end up with lots of abstraction
>> layers, which are good for some use cases, but bad for others: I see us
>> as a competitor to approaches like Blinx, see
>> http://www.w3.org/2008/10/24-mediaann-minutes.html#item01
>> who have just 5 categories for searching across *a lot* of videos.
>>
>>     
>>> Take the rationale about dates, in the MWG Guidlines (p.16). It
>>> introduces different *processes* in the lifetime of a digital image,
>>> such as *photo taken*, *image digitized* and *file modified*. Only after
>>> that can you discuss about the various date and argument that "date/time
>>> original" should be prefered over the other ones.
>>>   
>>>       
>> A good example: the description is very specific to metadata fields
>> specific to date / time, and does not argue with an overall conceptual
>> model. I think you can end up with a section like that by looking into
>> existing formats and their fields related to date / time, and
>> classifying these fields. Again: I don't think we disagree, but put a
>> different focus.
>>
>>
>>     
>>> My point is that, since this effort is necessary, it is better to make
>>> it explicit 
>>>       
>> Agree if we make it explicit *after* analyzing properties, or if we at
>> least start with the analysis, and not with thinking about the model.
>>
>>
>>     
>>> -- even if only as prose, rather than emerging only through
>>> its consequences in each property mapping.
>>>       
>> I see "only throughits consequences in each property mapping" as a
>> benefit. AFAIK the metadata WG deliverable was also written in such a
>> "bottom-up" approach, and the conceptul model you see is a result or
>> something emerging "on the way", and not the start of that work.
>>
>>     
>>>  I see the following
>>> advantages to this:
>>>
>>> - making the ontology/API easier to understand
>>> - enabling implementors of the API to extend it in a consisten way
>>>   - to properties that we would have considered out of scope
>>>   - to format that we would not have considered
>>>   
>>>       
>> I agree with all these advantages.
>>
>>     
>>>  
>>>       
>>>>> [example of "album title", "original album title"]
>>>>>       
>>>>>           
>>>> Good example. My opinion is exaclty that mapping these properties to
>>>> Dublin Core *does not* require finding these equivalences. From our
>>>> charter:
>>>> http://www.w3.org/2008/01/media-annotations-wg.html
>>>> Success criteria: "Abilitity to convert core metadata information from
>>>> one metadata standard to an other". This does not talk about
>>>> roundtripping, so information loss  is OK.
>>>> Out of scope: "Full coverage of all metadata elements in EXIF, IPTC,
>>>> XMP, MPEG-7, and similar broad vocabularies, is out of scope".
>>>> I interpret this as "it's OK to loose the structure of ID3 tags is lost
>>>> in the mapping to ID3.".
>>>>     
>>>>         
>>> Information loss is OK indeed, but we should be aware of the information
>>> we are losing.
>>>   
>>>       
>> I agree.
>>
>> Felix
>>
>>     
>
>
>   

Received on Tuesday, 25 November 2008 12:14:59 UTC