See also: IRC log
<tobias> We moved forward Wonsuk's demo presentation in the agenda.
<tobias> Wonsuks presents his demo of a reference implementation of the group's work
<tobias> Scope of the implementation: target formats are EXIF, MediaRSS, TV-Anytime
<tobias> Ontology is implemented in OWL (based on the recent version of the mapping table from may,19th)
<tobias> API is implemented using Jena, sanselan (EXIF libary) and Xalan
<tobias> scribe: tobias
One problem Wonsuk encountered: ma-elements might map to multiple locations in the standards
Example: keyword element in TV Anytime
Werner: we are on the level of fragments; some keywords apply to the whole resource, some to fragments
Wonsuk: Should we prioritize some elements which are returned?
Jean-Pierre: For TV-Anytime some
descriptions (if present) must replace others.
... If you use DC you have to pay the price of simplicity; many
elements have to be mapped, for instance, to dc:description
<joakim> agenda: F2F 4 Stockholm
Werner: We are still missing a
definition of data types for the properties
... Why is ordering of properties important?
Chris: How can you take care that you can elements which were joined can be (dis-)joined again.
Jean-Pierre: There are many problems with having a minimal set of properties, the mapping of more specific to more general properties is one of them.
Wonsuk presents his OWL ontology which models the mappings
Jean-Pierre: Why can't we use SKOS here? SKOS predefines some of the properties that Wonsuk defines in his ontology.
Wonsuk presents his implementaiton of the API (GET)
Wonsuk presents his implementaiton of the API (SET)
Werner: Do you already evaluate if the property is more general / more specific?
Wonsuk: not complete yet; because based on XPath
Florian: How can you deal with
changes in the standards on which the mappings are basedß
... can you support two versions of one standard in
parallel?
Now: Coffee Break
We are back from the coffee break.
Wonsuk presents open issues with respect to his implementation. These include adding types to the table, create a testdata set, exchange underlying libraries, defiine the scope of the SCOPE API
Tobias: library to parse EXIF files: http://sourceforge.net/projects/libexif
Wonsuk: Presents API proposal
Joakim: Jonas implemented an API for a metadata format recently.
Jonas: they specified a function for each property; so: a getter for each property
Wonsuk: API option 1: one getter
method for each property; option 2: handle each property via
input parameter
... Open issue: what kind of action could the API do with
semantic mapping like exact, related, etc?
Joakim: proposals occur more with the setters
David: For the getter you could return one property which has a more narrower meaning.
Intermediate Topic: How to handle the mapping table in the feature
Thierry: Excel table does not produce valid HTML output.
Thierry proposes to use the Wiki which can be transformed using XSL into a mapping table.
ACTION for Thierry: copy the format sheets into the Wiki.
<trackbot> Sorry, couldn't find user - for
ACTION for tmichel: Copy the format sheets into the Wiki.
<trackbot> Sorry, couldn't find user - for
<scribe> ACTION: tmichel to copy the format sheets into the wiki [recorded in http://www.w3.org/2009/06/25-mediaann-minutes.html#action01]
<trackbot> Created ACTION-129 - Copy the format sheets into the wiki [on Thierry Michel - due 2009-07-02].
<scribe> ACTION: tmichel to create an XSL transformation for the format mapping sheets [recorded in http://www.w3.org/2009/06/25-mediaann-minutes.html#action02]
<trackbot> Created ACTION-130 - Create an XSL transformation for the format mapping sheets [on Thierry Michel - due 2009-07-02].
One of the next steps: we have to work on the data type definitions
<scribe> ACTION: tmichel to add a datatype column to the format mapping sheets [recorded in http://www.w3.org/2009/06/25-mediaann-minutes.html#action03]
<trackbot> Created ACTION-131 - Add a datatype column to the format mapping sheets [on Thierry Michel - due 2009-07-02].
Wonsuk proposes to add XPath expressions to the sheets
Chris asks if having an XPath expresion is too specifc? Wonsuk says that he needs this for the implementation of the getters.
<victor> How many of the considered formats are XML-based?
Chris: In some metadata formats things can be expressed in different ways which makes having different XPATH expression necessary
<victor> What about non-XML based formats?
@Victor: most formats have XML serializations, but a considerable amount of them has also an RDF-based serialization
@Victor: EXIF for instance is non-XML based
Chris: thinks that the XPath expressions are just an implementation issue.
<scribe> ACTION: tmichel to add a column for XPath expressions to the format sheets [recorded in http://www.w3.org/2009/06/25-mediaann-minutes.html#action04]
<trackbot> Created ACTION-132 - Add a column for XPath expressions to the format sheets [on Thierry Michel - due 2009-07-02].
The group decided @ 12:15 that we should add XPath expressions to the mapping table.
Victor: What should we do if a format is not XML-based?
Werner: All formats that are non-XML based in our current mapping table are very flat.
<victor> "very flat" means that an offset in bytes is enough?
Current topic: check WIKI page of possible enhancements for a next revision of the ontology working draft
First point: Add a browseable version of the proposed ontology?
Werner: The idea was to have a structured definition of the elements.
Current point (from Werner): distinguish between descriptive and technical properties?
Joakim: the issue was: why should we include some technical properties and others not?
Chris: How did we come up with the list of properties in Barcelona?
Joakim: we voted in Barcelona
Chris: Should we try to get some feedback about the properties?
Werner: Discussed the properties in a Digital Cinema project
Jonas: why were we thinking of excluding technical metadata?
Werner: the idea was to keep the basic set quite small.
Jonas: OpenSocial has a lot more elements and are expanding all the time.
We might thinking of seperating between descriptive/technical metadata
Agenda topic: Lunch break
<dsinger> ACTION: tmichel to Copy the format sheets into the Wiki. [recorded in http://www.w3.org/2009/06/25-mediaann-minutes.html#action05]
<trackbot> Created ACTION-133 - Copy the format sheets into the Wiki. [on Thierry Michel - due 2009-07-02].
Agenda topic: Revision of core attributes
<wbailer> scribe: wbailer
splitting into core set of descriptive properties and technical metadata (maybe specific to media types)
david: propose to add frame rate
classify technical metadata according to media type
david: is frame rate and audio sampling rate the same? what to return for a movie?
jean-pierre: use edit rates for temporal sampling
joakim: call intrisic properties of media "technical metadata"
david: profiles are orthogonal to descriptive/technical metadata, eg. camera man only makes sense for video
define framerate (video) and samplingrate (audio) as properties
chris: describing properties of container vs properties of streams
jean-pierre: need to define which stream is main reference point of temporal addressing
chris: container with different bit rates
jean-pierre: containers with streams with different rates currently not an issue on the web
david: is bit rate needed or can
it be calculated from size and duration?
... add dataSize (file size or stream size)
... having bitrate additionally, it may be contradictory
... dataSize could be indefinite for live broadcasts
... use bit rate, if finite duration, data size could be
calculated
jean-pierre: which bit rate? average?
david: average bit rate
joakim: include sample type/size?
drop
joakim: number of tracks?
add number of track and track type
add getNamedFragments - possible fragment names to request
david how to get track format
david: accessibility issues for
multimedia
... requirements for adaptations, specific tracks may be
available for that purpose
... eg. high contrast video instead of normal video track
... would need to add function of track in addition to track
format
chris: mpeg-21 looked at
description of adaption, user needs
... not sure if adaptation requirements need to be added in the
metadata
... similar to age adaptation
tobias: has in his ontology number of views, price, and specific types of dates
<tobias> the dates are: creation date, modification date, publication date
<tobias> I also had more descriptive properties such as keywords, tags, category
david: can track type be retrieved using format on the track fragment?
number of views? dropped
chris: difference tag and keyword?
joakim: tag is user defined keyword
chris: keywords summarize, tags are subjective
tag will not be added
david: language versions of some
elements (eg title)
... what is returned?
... data types: uris vs text
... how to return sets of elements
... duration of file vs annotated duration
... unit of frame size, ... ?
... and (pixel) aspect ratio?
... how to specify compression format?
... no standard way of referring to compression type
... support gps location for notation, as well as named
places
tobias: why do we have location, but not person?
chris: can be depicted or shot location?
joakim: yes
chris: if we support depicted locations, we should also support persons, events, ...
joakim: we have a structured way
of expressing location, not for the other properties
... define in api whether you get depicted or shot location
chris: use location only for shot
location
... for depicted locations, persons, ... could be
keywords
... or add specific properties for these
<scribe> ACTION: dsinger to update description of location to represent shot location [recorded in http://www.w3.org/2009/06/25-mediaann-minutes.html#action06]
<trackbot> Sorry, couldn't find user - dsinger
<scribe> ACTION: david to revise e-mail on issues/updates of property description [recorded in http://www.w3.org/2009/06/25-mediaann-minutes.html#action07]
<trackbot> Created ACTION-134 - Revise e-mail on issues/updates of property description [on David Singer - due 2009-07-02].
david: difference between rating and target audience?
joakim: rating is user rating
david: is target audience a triple: issuer, target audience, value?
jean-pierre: triple: agency (optional), rating scheme name, value
<Daniel> hi all...sorry for missing the F2F, I am now in Seattle for another meeting
<Daniel> did you have a lunch ?
add price?
david: price for whom to whom?
tobias: also had license item
jonas: two versions of content: two separate resources
joakim: one resource for two instances
<dsinger> price - rental or own? what resolution? director's cut or the in-flight movie? and so on...
jonas: how to handle two technical descriptions?
joakim: does our definition still hold (describing abstract resource or instance)
chris: you do not describe technical properties when talking about the abstract resource
<Daniel> *
<Daniel> \
joakim: content star wars, 2 instances?
victor: at least 3 different descriptions in this case
simon: relations between these different descriptions are not yet defined
chris: we have a locator that points to an instance
simon: what about metadata of instance?
use relation element to point from instance to abstract content?
chris: use relation field
jonas: price more considered to instance
victor: price is described within a license
joakim: drop price
victor: dc:source would be more appropriate to link from instance to abstract resource
joakim: source was not even in top rated tags list
<dsinger> for frameSize, we have the problem of non-square pixels. Thus, I see the API could return (a) the actual pixel counts, ignoring pixel aspect ratio - useful for finding a 1080p movie, for example (b) the pixel counts, but adjusted to be the space they would occupy on a square pixel display if the more tightly-sampled axis is expanded (i.e. no resolution loss) - useful if you want to know how big a space, or a space of what aspect ratio, to reserve to display
<dsinger> on compression type, I recommend a return result that is the mime string of the container, with the codecs parameter as defined in RFC 4281, if needed
<victor> ...
<dsinger> for location, we decided it was actual shot location, so I suggest a pair, either of which elements can be empty, of a GPS location in ISO 6709 Annex H format, and a human ('annotated') name (e.g. "Barcelona", "Granny's house")
<victor> Werner: Before closing the discussion on metadata elements, we should review some of them. For example, "tagline"
<dsinger> title can have various sub-forms, e.g. sub-title, tagline, etc., as well as each possibly having language variants
<florian> http://www.w3.org/2008/WebVideo/Annotations/wiki/Candidate_Additional_Elements
<victor> see http://www.w3.org/2008/WebVideo/Annotations/wiki/Candidate_Additional_Elements
<victor> David: We need ma:relation
<victor> Joakim: Examples of subtypes could be: "thumbnail", "trailer", etc.
<victor> Following the candidate additional elements, we review the "rights" element:
<victor> Tobias. We have only a ma:copyright element
<victor> Victor: We had also intitially dc:rights but this element has been dropped....
<victor> ACTION Joakim Will refine the subtypes, like ma:contributor, ma:publisher etc.
<trackbot> Created ACTION-136 - Will refine the subtypes, like ma:contributor, ma:publisher etc. [on Joakim Söderberg - due 2009-07-02].
<victor> David: The mapping table in the wiki would be better located than in a spreadsheet
<victor> Tobias: Each element description should be in a different page in the wiki
<victor> Joakim: This would avoid having s many times the same text (what would be difficult to update etc.)
<victor> Werner, Wonsuk: nodds
<victor> Joakim: Draws in the whiteboard a scheme of the main page of the mapping table
<victor> Joakim: (drawing) Main page consists of a table with the following columns: MAWG, Description, DataType (example: "ma:contributor", "text description", "xsd:string")
<scribe> ACTION: wbailer to update attributes table [recorded in http://www.w3.org/2009/06/25-mediaann-minutes.html#action08]
<trackbot> Created ACTION-137 - Update attributes table [on Werner Bailer - due 2009-07-02].
<victor> Joakim (drawing): each mapping table (XMP, EXIF, etc) has also a table with the following columns:MAWG, XMP attribute, How to do the mapping, Datatype, XPath
<victor> Now, analyzing "List of fragments"
<victor> Joakim shows a Felix email, where he suggests the possible need to annotate his favourite fragments in a video
<victor> (It is discussed but a conclusion is not reached for the moment...)
<victor> David: Why not adding a new row: "fragment"?
<victor> Werner: agrees
<victor> Joakim: Does not like the term very much...
<victor> It is agreed that chapters are really useful
<victor> ma:fragment is proposed to be added
<victor> : chapter, highlight, favourite
<victor> as subtypes.
<victor> Chris: They are pairs <identification, label>
<victor> Agreed: ma:fragment is a pair of fragment, and the role it plays.
<victor> NEXT ITEM IN THE LIST of new elements to be considered: "Accesibility Metadata"
<victor> ACTION wbailer to update the candidate additional elements
<trackbot> Created ACTION-138 - Update the candidate additional elements [on Werner Bailer - due 2009-07-02].
<victor> NEXT
<victor> we review the list of improvements and enhancement of the ontology, see http://www.w3.org/2008/WebVideo/Annotations/wiki/Improvements_/_Enhancement_for_a_revision_of_the_"Ontology_for_Media_Resource_1.0"_document
<victor> David: I suggest adding a "Example" column in the mappings page in the wiki
<victor> Now: number 6 in the list of improvements: some in-scope formats are not part of the mapping table.
<victor> The list of "formats in scope" is in Clause 1.2 of the "Ontology for Media Resource 1.0" document
<victor> Joakim this can be fixed for the next version, no need to spend more time now on it.
<victor> Now: number 7-º0 in the list of improvements: we go over them very fast.
<victor> Now: number 11 in the list: Veronique said ontology definition etc. could be improved
<victor> We agree that all of it can be done for the next version of the ontology document
<victor> ACTION florian To check with Veronique (and Victor) the improvements 9-11 in the list.
<trackbot> Created ACTION-139 - Check with Veronique (and Victor) the improvements 9-11 in the list. [on Florian Stegmaier - due 2009-07-02].
<victor> (Jean Pierre Evain leaves the meeting to take a plane)
<dsinger> (Joakim goes with him, we note with fear the inward-opening locked door and the bars on the windows)
<Daniel> \me what happen in there ?
<dsinger> (Joakim is unable to re-enter; perhaps grateful)
<victor> We recall that some syntactic information (a column?) is missing in the mapping table
<victor> We recall that the next version might include several layers of complexity, and several layers of conformance
<victor> We postpone the meeting until tomorrow
<scribe> Meeting: MAWG F2F
<scribe> ACTION: chris to come up with examples [recorded in http://www.w3.org/2009/06/25-mediaann-minutes.html#action09]
<trackbot> Created ACTION-140 - Come up with examples [on Chris Poppe - due 2009-07-02].
This is scribe.perl Revision: 1.135 of Date: 2009/03/02 03:52:20 Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/ Guessing input format: RRSAgent_Text_Format (score 1.00) Succeeded: s/sso/s/ Found Scribe: tobias Inferring ScribeNick: tobias Found Scribe: wbailer Inferring ScribeNick: wbailer Scribes: tobias, wbailer ScribeNicks: tobias, wbailer Default Present: pchampin Present: David Jean-Pierre Tobias Chris Werner Florian Simon Jonas Anna Victor Azadeh Thierry Xonsuk Joakim Got date from IRC log name: 25 Jun 2009 Guessing minutes URL: http://www.w3.org/2009/06/25-mediaann-minutes.html People with action items: chris david dsinger tmichel wbailer[End of scribe.perl diagnostic output]