Re: comments on OMR 1.0 mapping to IPTC Schemas

Hi Michael,


2010/4/16 Michael Steidl (IPTC) <mdirector@iptc.org>

>  Hi Felix, see more below ...
>
>
>
> *From:* felix.sasaki@googlemail.com [mailto:felix.sasaki@googlemail.com] *On
> Behalf Of *Felix Sasaki
> *Sent:* Friday, April 16, 2010 2:00 PM
>
> *To:* Michael Steidl (IPTC)
> *Cc:* public-media-annotation@w3.org
> *Subject:* Re: comments on OMR 1.0 mapping to IPTC Schemas
>
>
>
> Hi Michael,
>
> 2010/4/16 Michael Steidl (IPTC) <mdirector@iptc.org>
>
> Hi Felix
>
> The issue I raised is about “what are you mapping”. To my understanding
> from reading the specs it is a mapping between metadata properties  - which
> are grouped by what? By their namespace? Or by an “existing format”, and
> what is a format – a named specification?
>
> The IPTC as a standardisation organisation would prefer that properties are
> considered as intellectual property of their makers. Thus a mapping is
> established between the ma: properties and the Dublin Core properties and
> not XMP which is only using Dublin Core properties but not having specified
> them. Further the mapping between ma:title and dc:title should be applicable
> regardless of the format which is used to annotate the Dublin Core Title
> property.
>
> To add: I understand what you are aiming at with the API below, but this
> would work in exactly the same way if the mapping is named “to Dublic Core”
> and not “to XMP”. So what I’m pointing at is more a naming and IP issue and
> not a technical issue.
>
>
> Understand. I tend to disagree that this is not a technical issue: if you
> do not enlist the potential field names which may occur in a metadata file,
> the API would need to check for everything before doing the mapping. So
> having the information available "looking for ma:title in XMP files, only
> search for dc:title" makes that easier.
>
> Besides, again on the technical level, XMP defines value types like LangAlt
> for dc:description, which are not available in Dublin Core itself. So having
> a place to gather these types is very for the API.
>
>
>
> OK, agreed, this is a feature beyond the Dublic Core specs.
>
>
> Your IP concerns are very important, and I am thinking of how to implement
> them while keeping such information . Would it help  if the table and
> column 3 would be renamed "Schemas used in XMP", and to add a statement to
> the specification like this:
>
> "XMP allows for using properties which are not from the XMP namespace, like
> Dublin Core. The specification of mappings between these properties and the
> Media Annotations Working Group vocabulary does not entail any intellectual
> property relation between the maintainers of these properties, e.g. the
> Dublin Core Metadata Initative, and XMP."
>
>  I think the Media Annotation group has to make a split regarding the API
> specs into features regarding the “annotation technology” (e.g. XMP) and the
> properties:
>

I agree to your three bullet points below, and I am currently reworking the
XMP mapping, see http://fabday.fh-potsdam.de/~sasaki/mawg/xmp.html . I am
1/3 through and have already included EXIF related metadata and photoshop as
well. Getting to keywords, I will take IPTC into account as well. So I hope
that this is the right way forward. If you have any specific comments,
please let me know :) . Btw., I will also ask Frank Biederich from Adobe
from feedback, probably later today when I have reached ma:numTracks.



> -          Exif metadata may also be expressed by the XMP annotation –
> won’t they benefit from the XMP features for Dublin Core?
>
> -          Adobe has defined some metadata in their own “photoshop”
> namespace (actually referring back the IPTC IIM metadata –
> www.iptc.org/IIM) like keyword. This would perfectly fit the ma:keyword -
> but is currently excluded from the XMP mapping.
>
> -          We, the IPTC, have defined a set of more refined Photo Metadata
> in the IPTC Extension namespace – which also builds on XMP.
>
> As XMP is an important annotation technology for visual content I would
> support that the API takes XMP features into account and provides e.g. a
> parameter to get a description in specific language. But the mapping from
> ma: properties to properties of other namespaces should not be strictly
> linked to the technology.
>
> Having a look into mediaont-api: you define interfaces for accessing
> properties, e.g.
>
> interface StringObject: MAObject, Language {
>
>                                    attribute DOMString value;
>
>                         };
>
> Perfect: you („only“) need a class for each annotation technology or
> format: the XMP class, the NewsML-G2 class ;-) etc. Each class implements
> the interface in a way which is specific to this annotation technology, then
> the user only has to now this file includes metadata in a specific
> annotation technology, then applying the right class would deliver what’s
> expected.
>

That sounds like a very good idea to me.

Best,

Felix


>
>
> Best regards
>
> Michael
>
>
>
>
> Best regards,
>
> Felix
>
>
>
>
> Br
>
> Michael
>
>
>
>
>
> *From:* felix.sasaki@googlemail.com [mailto:felix.sasaki@googlemail.com] *On
> Behalf Of *Felix Sasaki
> *Sent:* Friday, April 16, 2010 12:05 PM
>
>
> *To:* Michael Steidl (IPTC)
> *Cc:* public-media-annotation@w3.org
> *Subject:* Re: comments on OMR 1.0 mapping to IPTC Schemas
>
>
>
> Hello Michael,
>
>
>
> thank you for your very useful and detailed comments. I am not working on
> this mapping, but on the XMP mapping table, so I have a comment related to
> that below.
>
> 2010/4/15 Michael Steidl (IPTC) <mdirector@iptc.org>
>
> .....
>
> ------------
>
> In addition to this review of the IPTC mapping I would like to add one more
> comments:
>
>
>
> * as the IPTC is very involved in XMP we would like to point out, that the
> “Table 1: XMP” is misleading: the “XMP Attributes” like dc:contributor etc
> are definitely not part of the XMP specifications, as the namespace prefix
> indicates they are part of the Dublin Core specification. XMP is a
> completely metadata property agnostic framework based on RDF/XML – and Adobe
> makes only use of some properties in their implementation for Adobe
> products, like Photoshop etc. But it would be fully XMP compliant to have an
> XMP packet without a single Dublin Core property but Descriptions,
> Identifiers etc. from other namespaces.
>
>
> Sure. However, keep in mind that the ontology is supposed to be used to
> provide mapping relations for an API. Below is a sample method, adapted from
>
> http://lists.w3.org/Archives/Public/public-media-annotation/2010Apr/0058.html
>
>
> maobject1 = setMAResource ("video.flv");
> maobject1.getProperty("title");
>
> So you want to be able to get the title of a media resource - which is
> AFAICT stored as dc:title in XMP.
> Of course there might media files without dc:title, as you pointed out.
> Because of that I would map ma:title to let's say:
> - dc:title, exact mapping, dataype langalt (allows for language
> alternatives)
> - xmpDM:album, related mapping, album title in the XMP Dynamic Media schema
> - xmp:Nickname related mapping, text datatype, XMP Basic schema
> - xmp:Label, related mapping, text XMP Basic schema
>
> Also the XMP mapping table needs a thorough revision, and I am currently at
> it, with more updates later today.
>
> Best regards,
>
> Felix
>
>
>
>
> Many thanks for this work as we see that NewsML-G2 provides one of the most
> complete mappings to OMR of all referenced metadata schemas, so it may make
> sense to stay in touch.
>
>
>
> Best regards
>
>
>
> Michael
>
>
>
> ==================================================
>
> Sent by:
>
> Michael Steidl
>
> Managing Director of the IPTC <mdirector@iptc.org>
>
> International Press Telecommunications Council - http://www.iptc.org/
>
> Business office address:
>
> 20 Garrick Street, London WC2E 9BT, United Kingdom
>
> Registered in England, company no 101096
>
>
>
>
>
>
>