See also: IRC log
<scribe> scribe: various
<Daniel> our discussion focuses on what properties are requrired and mapping these with ontology
<Daniel> wonsuk: if you made an initial draft, please upload it to the wiki page
colm: we have developed an API since many
customers asked for it
... will demonstrate general and complex interaction and an demonstration purpose application
... we have 5 entities: reference to page, title like "obama vs. ..."
... .summary, content, ...
... document specific info like domain, channel (e.g. "you tube")
... tags from the youtube object itself
... language, media duration / width / heigth
... embedd tag, but not for all medias
... a publish date and a user friendly date (e.g. "9 days ago")
raphael: you have your own XML Schema for that?
... example with obama video, search query was "joe meets obama"
raphael: do you keep track of folksomonies in the metadata?
colm: no, we use only what is in the feed what we get.
raphael: most of values for this element are specific to your format, e.g. date is not an ISO date
colm: yes, we just developed our own format
raphael: if you need to exchange the metadata your need to tell people that you use not the ISO date
colm: we can make transformations in the API that solves that
... a more complex example: for each object we have a macro layer with information about the transaction, e..g languages that appear in the results
... they have their own categories, application specific ones, e.g. their own title / link / publish date / ...
... s/more complex example/more complex, application specific example/
... in the application they have a web page that makes use of these categories for searching
raphael: in the metadata you had English and "en"
colm: they did the merging, we put both in and it is
up to them to do the merging
... a more complex application: a video with transcription, and the user can click on the transcription and the audio of the video jumps to that passage
... the application can also do major scence detection
daniel: that is automatic speech recognition?
plh: how long does processing of a video take?
colm: 1:1, it is roughly a lot of plugins
which run in paralell
... and this is not a function of the resolution size
... for text video alignment we use ctm
raphael: the metadata comes from the schema you showed?
... we also can export the schema. CTM has (also) an XML serialization
silvia: the 5 top elements?
colm: datestring / date / summary / title /
reference, plus content which can have whatever you want
... e.g. media duration, media width, meida height, media format, media type string, embed tag
raphael: ontology is not exactly like an object model
... since you have inferences
... e.g. if you have a new class in an ontology, you might get changes in the model, because of the OWL semantics
... that is why you don't have "getter" and "setter" models for everything in the ontology, but different design patterns
... there are examples of removing a property in the ontology
... you cannot remove the relation because after reclassification a class might appear again
... an ontology has basic rules to be consistent
... you apply these basic rules when you change the ontology to make sure that the ontology is consistent again
... main design decisions of this API is to use the Command design pattern with a visitor pattern
... so you do not have a get / set method, but a visitor pattern. If you make a change then that might influence the whole ontology
... I will give some examples.
raphael: imagine we have an ontology as the output of our group. Somebody wants to add metadata.
raphael: dataFactory will give you individuals , from that you
built the triples
... AddAxioms might trigger more changes in the ontology model
veronique: so it is an API for changing ontologies, but it can also be used for querying instances
raphael: so in summary, you have inferences behind, and these make sure that the overall model is always consistent
colm: we have only one method, no set values
... we have a giant search application, that's it
... so what we have is a subportion of what people want to achieve here
capabilities are the second step
... one of the target is an API for browsers, they are not interested in "set"
raphael: for authoring, you might need the set capability
plh: will show the demo of the use
case I have in mind
... in my service I have an URI as an input, from that RDF is generated
... this uses an exif library, which generates the RDF.
... it extracts all kinds of metadata (EXIF, XMP, ...)
is easy to embedd metadata in an image, but not easy with a video
... what will our API do about that?
plh: that is where the ontology comes into
... I mean basically a set of terms
... if we have have a set of terms like "getAuthor", the browser can see what format is available and can get the information
... in this use case I assume that metadata is in the image itself
rapahel: for video that is much harder
felix: we had an proposal to link to external media information from the <video> tag
plh: HTML people currently prefer to look only to media internal information
silvia: I heard that the
external information will be discussed in a later step
... we should not restrict us with these options
plh shows API example, very simple, getTitle, getAuthor, getRights,
silvia: we had in mind to have only one method with a query string, so that it is extensible
plh: at the end we need both
silvia: at the end we need to identify the property name
larry: or constants to identify the property name
larry: adobe is very
interested in metadata
... XMP is very important for us
... XMP properties and media specific formats are being aligned, in a consortium called metadata working group
... they have published a deliverable on photo so far, video is to come
... they are concerned not so much on the web, but how to deal with metadata through production / assembly steps
... e.g. when a product is assembled form various parts: how do you take care of the meta data
... many people look only at the metadata of the final product
... we are taking steps before into account, including transcoding
... such capabilities are now being released as part of products which started shipping recently
... there is a transition between metadata and data, e.g. chapter marks which become part of the data on the DVD
... that is part of the processing changes, how to make the transition
... some of the specs of the metadata working group, e.g. metadata for video spec, are not public
... one level of the toolkit is public, additional utilities are not yet released, including some documentation
plh: did you work on XMP embedding in existing video formats?
difference in video formats about whether you can add metadata or not
... there are issues like one product is aware of XMP, another one only of EXIF
... the metadata WG has worked on how to deal with such "creative workflow" issue
silvia: we need to deal with both situations, final and "in process" metadata
larry: most metadata is a question of opinion
... most opinions change
... if you built semantic web way of metadata, keep in mind that provenance is here and might change
... history of changing and provenance of metadata are matters of fact
looking at XMP
larry: if you do external
metadata you need to care about syncronization
... the XMP version is new, tracks, frame rates, different kinds of annotation and markers were added
... they also analyze other formats
felix: if XMP is available is , is it always the first choice in provenance?
larry: look at the metadata consortium deliverable, the conflict resolution mechanism
raphael: if there are two sources of metadata which contradict each other, should not the application decide?
larry: consumers do not
want to decide, software should do that for them
... results are heuristics, results might be different for authors vs. time stamps
... a generic solution would still be complicated
... issue of merging metadata from different sources is very difficult
... seems to be more like a research project like a standards activity
... XMP has three parts: XMP model, some schemas which can be extended, and how metadata is embedded
... if you want a standard ontology look at part two
... most of the document is generic for any type of media
daniel: what is your opinion on our WG?
larrry: we support this activity, we encourage
you to go ahead
... we were not able to submit XMP so far
... technically there should be no reason for you not to use XMP
felix: so we could just use part 2 of XMP here?
larry: yes, but I need to make sure
plh: need to be aware that XMP refers to various other formats like dublin core
larry: XMP takes various parts of dublin core
plh: one way would be to take the core of XMP part 2, use if for the ontology and the API, and we are done
larry: there is a c++ library for reading and manipulating XMP
... and there is action script based separate model
... so that you can built standalone applications
plh: you did not try to align XMP to other metadata?
larry: that is part of the metadata consortium work
... and it is part of part 3 of XMP
silvia: still confused about legal side . Is XMP copyright by adobe or also metadata WG?
larry: not sure if that is a problem
... for a W3C member submission
plh: with a member submission you give your
copyright to W3C
... we might ask Adobe to give up parts of part 2
... W3C is not interested in all parts of XMP, e.g. "how to use XMP in flash"
... so adobe has the problem that in that case XMP would need to be splitted across organizations
larry: you could make a normative reference to (parts of) XMP
... I would not spend a lot of time on the license issue
doug: currently no standard for getting width
... I wanted an API for just that information
... and have developed a spec for that
... we want to be able to get at information, two kinds of metadata: intrinsic and extrinsic meta data
... e.g. width and height is intrinsic, but the creator is extrinsic
<chaals> http://dev.w3.org/2006/webapi/MediaObject/publish/ -> draft from end of last year
doug: we need authors need this for inserting
information in web pages or more complex information
... having the browser vendors early on will allow very early feedback
... otherwise browser vendors will not implement your work
... the draft above is a very generic way of getting any meta data
... so just a method "getMetadata"
... there should be some mapping between keywords
... working on API first and later ontology is important
larry: I think there is a real potential of mismatch in cross discussion
... if you think intrinsic properties like width / height, vs. descriptitve information like transcription
... I am not sure if the same API is useful for both
... there is the problem of having different vocabularies, e.g. resolution in terms of width / heigth, versus ratio
... media is updated independently of people who access it
doug: in terms of conflating intrinsic and extrinsic:
... it is very different to say "who is the author?"
... but if you make a generic API it does not matter what kind of information you have
... the line between intrinsic and extrinsic is blurry, see e.g. EXIF information
felix: is the browser perspective brought in by MS in the metadata WG?
larry: not sure
discussion on how much the API should differentiate between various kinds of metadata
larry: usage of namespaces in XMP might be a problem for querying XMP metadata
doug: I was mainly concerned with a web browser API
larry: what are you use cases?
... examples of web applications that want to use your API
doug: see flickr, it would be nice to
put tags into a picture
... if I have embedded captioning
renato presents PLING
Felix: currently we are considering having just a slot for expressing rights
renato: understand, when you need some more information about policies in your framework, please come back to us. We are looking at general frameworks for policy languages
you want to describe "who has already seen this?" you might run into policy issues which are
addressed by PLING
... you (media annotations WG) provides answers on "how is meta data attached to video?", PLING is concerned with the question "how to attach policy information?"
... we created PLING because many companies do not know how to use various languages together
... in the outcome of the media annotation WG we need a way to use an URI to reference to a policy description
looking at rights management schema from XMP
XMP spec, page 31
raphael: why two field for machine readable and human readable fields?
... the user agent can choose to follow the link and choose what he wants
rigo: two options: have a type argument with the content type, or two fields
jan: you need to give the author the freedom
... you want to be able to attach rights to e.g. media you are distributing
discussion on wheter a mechanism specifically for attaching a policy to medias needs to introduced or not
raphael: what is the main interest of Samsung?
daniel: technically several interests
... resolution for watching web videos on TV is no good
... comparing PC and laptop, TV has no good interaction between user and TV
... no scrolling or mouse
... so the question is how to make the video on the web useful for TV
... we think about a video recommendation system
... that can help users to find videos
... having a common ontology will help search and recommendation
... we are used to have a bookmark tool on the TV. But having an URI based mechanism (from the fragment WG) will be helpful for us
wonsuk: requirements doc is not finished yet
... mainly basic sceleton of the document
... as result of yesterdays discussion
wonsuk describes the document
wonsuk: how to link between sec. 2 and sec. 3?
veronique: subsections of sec. 3 can contain tables with links to sec. 2
schedule plan: veronique will edit the document, will send it around within the WG for another internal review
<Daniel> before official publication, working group will have recurculation process for reviewing and modifying the existing document. 2 weeks or a bit longer
felix will send XML later to veronique
<scribe> ACTION: Felix to get CVS accounts for veronique and wonsuk [recorded in http://www.w3.org/2008/10/24-mediaann-minutes.html#action01]
<trackbot> Created ACTION-24 - Get CVS accounts for veronique and wonsuk [on Felix Sasaki - due 2008-10-31].
<Daniel> requirements section willl be subsection of each use case section. e.g., Video Use Case (Req1, Req2, Req3, etc...)
<plh> ACTION: Felix to review XMP basic schema [recorded in http://www.w3.org/2008/10/24-mediaann-minutes.html#action02]
<trackbot> Created ACTION-25 - Review XMP basic schema [on Felix Sasaki - due 2008-10-31].
<plh> ACTION: Thierry to review XMP Dublin Core schema [recorded in http://www.w3.org/2008/10/24-mediaann-minutes.html#action03]
<trackbot> Created ACTION-26 - Review XMP Dublin Core schema [on Thierry Michel - due 2008-10-31].
<plh> ACTION: Felix to review XMP Rights Management schema [recorded in http://www.w3.org/2008/10/24-mediaann-minutes.html#action04]
<trackbot> Created ACTION-27 - Review XMP Rights Management schema [on Felix Sasaki - due 2008-10-31].
<plh> ACTION: Veronique to review XMP Media Management schema [recorded in http://www.w3.org/2008/10/24-mediaann-minutes.html#action05]
<trackbot> Sorry, couldn't find user - Veronique
<plh> trackbot-ng, status
<plh> ACTION: Véronique to review XMP Media Management schema [recorded in http://www.w3.org/2008/10/24-mediaann-minutes.html#action06]
<trackbot> Created ACTION-28 - Review XMP Media Management schema [on Véronique Malaisé - due 2008-10-31].
<plh> ACTION: Felix to tell Wonsuk to review XMP Basic Job Ticket schema and Paged-Text schema [recorded in http://www.w3.org/2008/10/24-mediaann-minutes.html#action07]
<trackbot> Created ACTION-29 - Tell Wonsuk to review XMP Basic Job Ticket schema and Paged-Text schema [on Felix Sasaki - due 2008-10-31].
<scribe> ACTION: wonsuk to review XMP Basic Job Ticket schema and Paged-Text schema [recorded in http://www.w3.org/2008/10/24-mediaann-minutes.html#action08]
<trackbot> Sorry, couldn't find user - wonsuk
<plh> trackbot-ng, reload
<plh> trackbot-ng, status
<plh> ACTION: Joakim to review XMP Dynamic Media schema [recorded in http://www.w3.org/2008/10/24-mediaann-minutes.html#action09]
<trackbot> Created ACTION-30 - Review XMP Dynamic Media schema [on Joakim Söderberg - due 2008-10-31].
raphael: tom baker should be the person to contact
<scribe> ACTION: Felix to contact tom baker about dc liaison [recorded in http://www.w3.org/2008/10/24-mediaann-minutes.html#action10]
<trackbot> Created ACTION-31 - Contact tom baker about dc liaison [on Felix Sasaki - due 2008-10-31].
<scribe> ACTION: Daniel to make a liaison with MPEG using information from the XP homepage [recorded in http://www.w3.org/2008/10/24-mediaann-minutes.html#action11]
<trackbot> Sorry, couldn't find user - Daniel
<scribe> ACTION: Soohong to make a liaison with MPEG using information from the XP homepage [recorded in http://www.w3.org/2008/10/24-mediaann-minutes.html#action12]
<trackbot> Created ACTION-32 - Make a liaison with MPEG using information from the XP homepage [on Soohong Daniel Park - due 2008-10-31].
<raphael> XG MPEG Liaison document: http://www.w3.org/2005/Incubator/mmsem/liaison/29n79301.doc
<raphael> More generally, go to http://www.w3.org/2005/Incubator/mmsem/wiki/Liaison
raphael: for IPTC, they are following our work, we will get informal comments
EBU covered by Jean-Pierre
<scribe> ACTION: Felix to evaluate if contact to IPTV Japan is valuable [recorded in http://www.w3.org/2008/10/24-mediaann-minutes.html#action13]
<trackbot> Created ACTION-33 - Evaluate if contact to IPTV Japan is valuable [on Felix Sasaki - due 2008-10-31].
<scribe> ACTION: Soohong to contact Open IPTV forum [recorded in http://www.w3.org/2008/10/24-mediaann-minutes.html#action14]
<trackbot> Created ACTION-34 - Contact Open IPTV forum [on Soohong Daniel Park - due 2008-10-31].
<scribe> ACTION: Joakim to check contacts to OMA [recorded in http://www.w3.org/2008/10/24-mediaann-minutes.html#action15]
<trackbot> Created ACTION-35 - Check contacts to OMA [on Joakim Söderberg - due 2008-10-31].
<trackbot> ACTION-16 Confirm that there should be a presentation about IMM closed
<trackbot> ACTION-18 And others to elaborate on the top-down approach about use case closed
<trackbot> ACTION-19 And others to look on the draft agenda for the TPAC meeting closed