MAWG - TPAC 2008 f2f

23 Oct 2008


See also: IRC log


Daniel, Wonsuk, Joakim, Veronique, Frank, Fabio, Erik, Davy, Felix, Silvia, Colm, Frank_(Canon), Herve, Daniel_(Expway), Youenn, Joerg, Karen, guillaume
Daniel, Felix
fsasaki, raphael, Felix, erik


meeting start


<scribe> scribe: raphael

Introduction of the participants

Daniel: Chair of the group, work in Samsung, first time participation in W3C activities
... co-chair an internet protocol WG in IETF

Wunsok: Research Institute in Korea

Joakim: Erikson (operating system for mobile phones), interest in IPTV
... background in image classification, indexing and retrieval

Veronique: Vrij Universiteit in Amsterdam, work in the Cultural Heritage Domain

Frank: assistant professor in UvA in Amsterdam, worked in the MPEG-7 WG (responsible for DDL)

Fabio: Paris, Thomson, interest in video over P2P

Erik: project manager in IBBT in Ghent (Belgium), co-chair of the Media Fragment WG, work with broadcasters and in the cultural heritage domain

Davy: researcher in IBBT Ghent, multimedia annotation

<fsasaki> Raphael: was co-chair of MMSEM XG, now co-chair of media-fragments WG. Interested on NLP apllications and multimedia.

<fsasaki> .. used to work previously in cultural domain setting

Guillaume: South Africa, interest in languages, content management systems ... would like to consider videos as real structured documents

Felix: W3C staff contact, background in using XML-based or RDF-based languages for representing linguistics resources
... will leave W3C in March, there will be another staff contact

Silvia: run her own company in Australia, did Annodex, CMML, TemporalURI, work with Mozilla on accessibility of Video, active in the Open Source community

Colm Doyle: technical director of Blinx, a video company

Frank: observer from Canon research in France, what could be the link with the Media Fragment WG

Herve: also observer from Canon

David: Expway, expert in binarization of XML data, interest in media annotations

Johan: observer from Canon, want to get an idea of what is going on here

Daniel: going through the charter and the goals of the group

charter review

Daniel: goals is to provide a minimal ontology for representing metadata + to provide an API for accessing video metadata
... timeline is sharp
... goal of this f2f is to have quickly first draft documents

Felix: there will be a lot of missing pieces, this is the purpose
... we want to get quickly feedback from a wider community

Daniel: shows the wiki page
... goes through the Use Case & Requirements wiki page
... we will have another F2F meeting, in Ghent (Belgium) on 9 and 10 December

Felix: 2 more people join

Heuir: AC Rep of Siemens, Munich (DE), I was active in MPEG for descibing content of audio-visual content

Karen Myers: W3C

<fsasaki> scribe: Felix

MMSEM presentation

<fsasaki> slides at http://www.w3.org/2005/Incubator/mmsem/talks/MUSE2007/

<fsasaki> joakim: looking at liaisons - who is member of I3A?

<fsasaki> raphael: adobe, canon, many camera people

<fsasaki> .. similiar membership as IPTC

<fsasaki> .. difference between I3A and IPTC metadata is that in I3A there is four blocks of meta data

<fsasaki> .. problem is that I3A (DIG35 metadata) is licensed, we had asked for tools for processing it, but there isn't

<fsasaki> ACTION: Felix to add a link from WG page to the "still alive" MMSEM wiki pages and to make sure that everybody can edit the pages [recorded in http://www.w3.org/2008/10/23-mediaann-minutes.html#action01]

<trackbot> Created ACTION-20 - Add a link from WG page to the \"still alive\" MMSEM wiki pages and to make sure that everybody can edit the pages [on Felix Sasaki - due 2008-10-30].

<scribe> scribenick: erik

questionnaire on formats in scope / out of scope

<fsasaki> scribe: erik

felix goes over questionnaire in-scope vs out-of-scope

joakim: linking between standards is desirable

joakim/guillaume: is thumbnail data or metadata? it depends on UC

daniel/raphael: what is real definition of "media" in this WG?

daniel: primarily img, audio, video from questionnaire

felix: methodology of interrelating existing ontologies might be good way to look at things

veronique: maybe start from UC and find out what standards we need to describe

frank: what do we really want? ... not excluding media (cfr. exif)

daniel: focus should be video

raphael: "temporal media" covers better the scope

felix: discussion formats in-scope vs. out-of-scope should be video central (but the it should be looked at format by format if there's a good link)

erik: within Media Fragments WG, video is key (together with audio, image & timed text)

frank: even broadening and maybe considering haptic media too

raphael: motto keep it simple ... thus parts of MPEG-7 too low-level (color histograms)
... try to see overlap between all standards, then you probably already have some top-level stuff to start with

consensus out-of-scope: MPEG-21, SVG, SMIL

presentations on specific formats

<fsasaki> raphael presenting on IPTC photo meta data

<fsasaki> scribe: fsasaki

raphael: white paper IPTC photo metadata white paper in 2007
... IPTC photo metadata become a standard recently
... white paper authored by photo / image industry
... blocks: format descriptive, administrative, rights, technical
... each block points to some properties
... e.g. to EXIF

raphael describes some properties

raphael: properties have names or controlled vocabularies
... e.g. for location they have geo names URI
... for person they propose controlled vocabularies, e.g. wikipedia
... in administrative: lots of things from EXIF, location from others
... rights metadata: they re-use plus
... for technical properties: EXIF + some additional properties
... physical type of original is enumerated list

joakim: how about different video formats?

raphael: it is for still images, not video

rapahael: the white paper was from 2007. Open issue: how to identify images?
... they use URIs to identify images
... they developed their own URI (HTTP URI) based scheme, not registered with IANA, but they have registered other schemes
... all metadata is embedded in a file, e.g. header of JPEG file
... currently anybody can modify the metadata
... so they don't know if somebody has modified the data
... it happens often in the image industry, that authors change the date of an image to have it published
... so the publisher want to have waterproof techniques to avoid these problems

frank: a field said "maxium resultion is ..."
... how do they handle cases where the resultion does not fit?

raphael: currently they copy the metadata, so they have the danger of inconsitencies

going through the fields

no headline field

description - as a general field

location, person, event

raphael: encourage users to use a controlled vocabulary, say "this SHOULD be used", not with a MUST

guillaume: would be good to separate description and caption, also in the video

raphael: description is general

Colm: there is captioning, close captioning, summary, OCRing as another track, logos, facial recognition etc.
... these all needs indiviudal tags

guillaume: hard to derive properties from these, better to describe properties from this

Joakim, Raphael: better to have general properties that can be specialized

frank: who, what, when, why, where are important 5 "w"

raphael: for some properties there might be multiple occurences, that is different cardinality

there are different kinds of "locations", agreement that a location field is necessary, still need to decide whether one field, 2, more, ...

and how (hierachically) the properties should be structured

<raphael> distinguish cataloging properties versus descriptive properties

<raphael> ... thus not that bad to have different who, where, when, etc. for where the media has been captured, versus what it depicts

raphael: "who, when, what, where"
... see dublin core

<raphael> Dublin Core new Task Metadadata: http://dublincore.org/kernelwiki/FrontPage?action=AttachFile&do=get&target=KernelMetadataERCApplicationProfiles1_4a.htm

<erik> s/metadadata/metadata

veronique: should properties here be also linked together in this group?

daniel: after verifying relationship between standards we can have a common ontology in W3C
... but we cannot have much more

<raphael> +1 for V�ronique: this is most likely the hardest issue this group will have to solve

felix: maybe not even mapping, just describe what is commonly used

<raphael> I would like to have relationships between the properties ... but add serious complexity

daniel: assuming that metadata type A has a "film" field, and type B has a "movie" field. We want to be able to work across these, right?

<raphael> +1 for Guillaume: modeling provenance of the metadata is important too

guillaume: do we want to document where metadata came from?

example in terms of API: having a method "getLength" which gives you back the length and the origin (type) of metadata

<Karen> Felix reviews document from Metadata Working Group

<raphael> I'm not sure we should go into a conflict resolution mechanism

Felix: should we also go for such a conflict resolution mechanism as in the mwp guidance doc?

<raphael> I would prefer to have the ontology modeling all the necessary provenance information ... and let the application deals with that

agreement to have no such conflict resolution mechanism

Top-Down Modelling Approach

<guillaume> http://www.w3.org/2008/WebVideo/Annotations/wiki/Top-Down_Modelling_Approach

Felix: is this an alternative approach to the bottom-up approach?

frank: no, we should do both at the same time

frank continues presentation on "top down" appraoch

veronique: search and annotation are two parts of the same coin

frank: true
... starting "top down" will help us to get the highlights we need
... proposal is to achieve a minimal set of properties by looking at use cases, overlap when, see what standards have, is that what we need, and so we link to them

guillaume: do we first need to make a list of tasks we want to support?

frank: I think such a list would help to set priorities

joakim: the "browse" task is different from "search"

veronique: search and browse is like accessing the data

frank: analyze is in the direction of "working with material"

felix: propose to take the material in http://www.w3.org/2008/WebVideo/Annotations/wiki/Top-Down_Modelling_Approach in the requirements document into account

veronique: link between different tasks and existing standards can be a way to describe fields of different standards

agreement to put the material into the requirements document

<Daniel> going through our use cases one by one and editing by wiki pages

<Daniel> Cultural Heritage UC

<Daniel> not only audio and video for the cultural heritage, but also we should think about others media types ?

<Daniel> Felix is drawing on the board

<Daniel> UC: http://www.w3.org/2008/WebVideo/Annotations/wiki/CulturalHeritageUC

<Daniel> getting a data for annotation voca. how to use and change it can be parts of API draft

<FD> Felix: Example of a get API to get date information: getDate(URI, vocabulary)

<FD> Felix: vocabulary parameter could be "EXIF" for example

<Daniel> just moving UC discussion forward, and consider API issue later

<Daniel> coming back to the UC review...

<Daniel> some of low lever description can be parts of requirements...ex, interop. search, deploymet, fragment. etc

<Daniel> fragment parts may belong to another group (e.g., media fragment WG)

<Daniel> where to store URI and fragmented information ?

<Daniel> next UC: http://www.w3.org/2008/WebVideo/Annotations/wiki/VideoUC

discussing "The goal of this WG is to clean up this jungle and make our ontology support of commonly used properties for describing video content."

<Daniel> changing the above text into "the bridging of commonly used properties for..."

<scribe> ACTION: guillaume to write a use case "interaction, navigation" [recorded in http://www.w3.org/2008/10/23-mediaann-minutes.html#action02]

<trackbot> Sorry, couldn't find user - guillaume

<Daniel> newly adding "We do not aim to solve the semantic mismatch problem but leave it to the schemes which are used for annotation/retrievel" at the bottom of Example

<Daniel> new text proposed: We do not aim to solve the semantic mismatch problem but leave that to the application who creates the annotation / retrieval.

<Daniel> mapping between the different metadata may take palce in the Ontology document of the WG later.

<Daniel> UC: http://www.w3.org/2008/WebVideo/Annotations/wiki/MobileUC

<Daniel> Mobile use case already widely spreaded in the world...

<Daniel> geolocation info when contents generated...and use location are both valuable aspects.

<Daniel> geolocation in the web, for example: life logging...

<raphael> Raphael: I would point to a liaison with http://www.w3.org/2008/geolocation/

raphael, good point

<Daniel> looking at Geolocation WG charter and homepage on the screen

<scribe> ACTION: Felix to create a liaison to Geolocation WG [recorded in http://www.w3.org/2008/10/23-mediaann-minutes.html#action03]

<trackbot> Created ACTION-21 - Create a liaison to Geolocation WG [on Felix Sasaki - due 2008-10-30].

<Daniel> discussion: Support media adaptation for mobile device capabilities such as bandwidth, physical screen, audio and text. Media adaptation depending on business models and user preferences.

<Daniel> targeting for describing mobile device characteristics (screen size, codec style, etc)...in scope or out of scope ?

<Daniel> related activities are in OMA (Open Mobile Alliance) today

<Daniel> how to describe device characteristics is pretty different from media scope

<Daniel> dinner tonight together...:-)

<Daniel> PLING joint meeting tomorrow about Multimedia sharing

<Daniel> Mobile TV is deleted...

<Daniel> also, Geographic Location/Privacy WG in IETF: http://www.ietf.org/html.charters/geopriv-charter.html

<Daniel> resume: UC: http://www.w3.org/2008/WebVideo/Annotations/wiki/MultimediaPresentationUC

<Daniel> 7:30 .at lobby for dinner tonight

s/40 /30 /

<Daniel> adding a new sentence: "The WG is likely not to concentrate on this use case in the beginning but might come back to it later"

<Daniel> UC: http://www.w3.org/2008/WebVideo/Annotations/wiki/RecommendUC

<Daniel> going through UC

Daniel: amazon provides a service where users can search through data aggregated from different service
... such a service is similar to what we want to achieve with this UC

<Daniel> changing the title of UC: Recommendation across different media type

Daniel: title of UC is better "Recommendation across different media types "
... for this UC metadata descriptions need to be unified

Joakim: AMG has a lot of metadata as a relational database, about movies, music, games, which is interrelated

wonsuk: searching videos in youtube, you get recommendations based on the video you choose
... our ontology could help that youtube could provide yahoo and other videos

<Daniel> new UC discussion: http://www.w3.org/2008/WebVideo/Annotations/wiki/MultimediaSearchUC

<Daniel> question: leave it out or keep as it is...there are several relative parts in others UCs

agreement to cover this UC not seperatly but cover it as part of other UC and the "top down" section

<Daniel> new UC: multimedia sharing

<Daniel> no initial text for this UC

agreement not to cover this UC seperately

<Daniel> new UC discussion: http://www.w3.org/2008/WebVideo/Annotations/wiki/PhotoUC

agreement to take this as an input to our main UC video into account

<Daniel> new UC discussion: http://www.w3.org/2008/WebVideo/Annotations/wiki/MusicUC

<Daniel> changed the title into *Audio*

agreement to take this as an input to our main UC video into account

<Daniel> new UC discussion: http://www.w3.org/2008/WebVideo/Annotations/wiki/NewsUC

<raphael> +1 for music -> audio

<Daniel> Veronique: moving it to the relared UC (e.g., museum or others...) to make broader scope

<raphael> news is a domain specific use case ... not sure it should stand

<raphael> ... or might be like cultural heritage UC

<Daniel> TimeText for karaoke service on the TV, Mobile Phone

<Daniel> +1 for raphael suggestion

Felix: fine with having, it, just propose to take "XBRL" out of it

<Daniel> new UC discussion: http://www.w3.org/2008/WebVideo/Annotations/wiki/TaggingUC

<Daniel> I am away for a while to glance at AC discussion (future W3C structure) since I am a AC of my affilition...sorry...

veronique: we had said that we just provide a comon means to add / query tags, across formats, but not tackle the interoperabilty between tags

agreement within the group



agreement that we need a means to have the metadata both in the media and outside

requirement that we need to register a media type for external meta data


<raphael> do you mean a registration at iana ?

<raphael> w3c hates that .... though they have a precedent with xpointer

<raphael> most of W3C folks dislike the idea

raphael, yes, at iana. I have done that before and I'm still alive, and Philippe is the contact to IETF


discussing http://lists.w3.org/Archives/Public/public-media-annotation/2008Sep/0085.html

we will discuss this with raphael in the room

discussing http://www.w3.org/Bugs/Public/show_bug.cgi?id=6066

put it on the list of requirements, see later if we have people pushing for this

<scribe> ACTION: Felix to go back to Karen and check about " IPTV metadata specification" - what is it, is it available for us? [recorded in http://www.w3.org/2008/10/23-mediaann-minutes.html#action04]

<trackbot> Created ACTION-22 - Go back to Karen and check about \" IPTV metadata specification\" - what is it, is it available for us? [on Felix Sasaki - due 2008-10-30].

going back to http://lists.w3.org/Archives/Public/public-media-annotation/2008Sep/0085.html

joakim: distinction between abstract works and instances is good
... not sure what "manifestation" is

raphael: example for work is "work from S.B. Bach". Expression is "particular performance by an orchestra"
... manifestation is "recording in a particular year". Item is "this CD"
... we can have this model for describing things, but do we have use cases for it?

veronique: we have only two levels - manifestation and expression
... we can have just one layer but allow people to have a specialization for using their scheme

discussion on usage of various models, their (wide?) usage

so no final resolution yet about this requirement - but we will put it in the WD and say "we are not sure about this requirement yet" and also ask the public for feedback

above is issue 6130

veronique: maybe specific to cultural heritage use case
... and have this requirement as something we tackle after the first version of the ontology is done

requirements document timeline

on http://www.w3.org/Bugs/Public/show_bug.cgi?id=6113 again: proposal is to have for each property just a set of getter and setter function. Question: how does that relate to (hierarchy of) properties in the ontology

having initial version of the document in the next days, around a week for review within the working group, and publication of first WD after

<Daniel> rssagent, draft minutes

<scribe> ACTION: Felix to explain XMLSPEC to wonsuk [recorded in http://www.w3.org/2008/10/23-mediaann-minutes.html#action05]

<trackbot> Created ACTION-23 - Explain XMLSPEC to wonsuk [on Felix Sasaki - due 2008-10-30].

adjourned for today

Summary of Action Items

[NEW] ACTION: Felix to add a link from WG page to the "still alive" MMSEM wiki pages and to make sure that everybody can edit the pages [recorded in http://www.w3.org/2008/10/23-mediaann-minutes.html#action01]
[NEW] ACTION: Felix to create a liaison to Geolocation WG [recorded in http://www.w3.org/2008/10/23-mediaann-minutes.html#action03]
[NEW] ACTION: Felix to explain XMLSPEC to wonsuk [recorded in http://www.w3.org/2008/10/23-mediaann-minutes.html#action05]
[NEW] ACTION: Felix to go back to Karen and check about " IPTV metadata specification" - what is it, is it available for us? [recorded in http://www.w3.org/2008/10/23-mediaann-minutes.html#action04]
[NEW] ACTION: guillaume to write a use case "interaction, navigation" [recorded in http://www.w3.org/2008/10/23-mediaann-minutes.html#action02]
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.133 (CVS log)
$Date: 2008/10/23 16:04:00 $