MMSEM Athens F2F Meeting, December 8-9 2006 -- 8 Dec 2006

Good afternoon everybody

Jeff open the meeting

We have problems to connect to Zakim ...

<scribe> Scribe: Raphael

Jeff talk about the possibilities in the future, and mention the opportunity of having a WG

scribe: smooth transition between an XG to a WG
... will be a topic of discussion tomorrow
... the previous XG (content labeling) is trying to become a WG

I'm trying to scribe

1. Music Recommendation Use Case

Oscar Celma is

an official new invited expert

Introduction: personal music collections have grown

Music information Place: Audio (music recording), Text (lyrics), Video (video clips, CD covers)

Song metadata includes editorial, acoustic and cultural information

Users are represented by their profiles including demographic, geographic and psychographic information

Use case

Metadata initiatives for describing Audio Content

ID3 v2, OGG, APE v2, MusicBrainz 2.1 (RDF data not yet supported)

see vocabularies page

ID3, OGG and APE are tag based (free attribute value pairs)

All these approaches focus on editorial information

Audio Ontologies (in OWL)

Kanzaki, Music Production, Music Recommendation

See vocabularies

OGG Vorbis example

Interoperability problems and solutions:

scribe: Two general approaches: 1/ integration at RDF level (song, artist instances) ; 2/ ontology mapping ?
... but there is no really ontologies

Concrete example, see code in the wiki

Idea: take the commonalities between all the proposed tags by these schema (ID3, OGG, etc.)

Vassilis: for which reasons the ontology has been made ?

ID3 tags are inside the audio file

Jeff: is there some conflicting name between the various schemas ?

Advantages: possible scenarios ?

Create personalized playlists ... exploit song metadata (RDF), plus user profile

Upcoming gigs (user FOAF profile, geonames, RDF data about the venues and events

Simplify navigation in music collections through personalized views

integrate RDF music data into /facet

EM is Erik Mannens

Semi-automatic annotation of music

(link with the tagging use case)

Semantic podcast

Mention a wikilyrics that could have a webservice for getting the lyricsof a song

Paola: relation between a consumer and buisness; why not B2B ?

Oscar: it is possible

Future Plans:

search for commonalities between the various schema, build under the Music Recommendation ontology

Do a ID3 to RDF translation

<scribe> ACTION: Oscar and Jacco to provide a first implementation by Christmas [recorded in http://www.w3.org/2006/12/08-mmsem-minutes.html#action01]

<scribe> ACTION: Paola to provide a possible B2B scenario [recorded in http://www.w3.org/2006/12/08-mmsem-minutes.html#action02]

2. News Use Case (Thierry Declerk)

<mhausenblas> ScribeNick: mhausenblas

Raphael: General introduction, NewsML, XBRL. UC will focus on financial domain.

Thierry explains the 'Metadata Labels

scribe: TITLE tag contains a list (even sentence) of keywords
... not sure if keywords stem from a controlled vocabulary

<raphael> Translating XBRL into Description Logic

<raphael> Information about companie, profit and loss, etc.

<scribe> ScribeNick: mhausenblas

<raphael> Possible application: expose NewsML information in RDF

Raphael: Combine (lightweight) NewsML with XBRL descriptions

<raphael> link them with RDF version of XBRL and DownJonesML

<raphael> B2B scenario

<raphael> Challenge: merging all the data

Jeff: Concrete Examples?

Raphael: Hopefully 01/2007 real world data should be available
... Chris HW is in touch with DowJones
... Any contribution is welcome. Michael will try to get some real world examples from NM2

<scribe> ACTION: Raphael, Therry, and Chris to sync UC Wiki description with their new focus (financial domain) [recorded in http://www.w3.org/2006/12/08-mmsem-minutes.html#action03]

Paola: How about using Reuters test corpus?

<scribe> ACTION: Paola to supply XG with reference to corpus and usage license till next week. [recorded in http://www.w3.org/2006/12/08-mmsem-minutes.html#action04]

<raphael> Back after the coffee break

3. MPEG-7 Use Case (Stamatia)

<raphael> Chrisa Tsinaraki will join this use case

<raphael> Motivation: MPEG-7 is a set of XML schemas

<raphael> ... lack of explicit and precise semantics

<raphael> ... effort to attach formal semantics to mpeg7 (ontologies)

<raphael> Example 1: semantically identically metadata can be represented in multiple ways

<raphael> Example 2: image representation in itself as a multimedia entity can be performed in multiple ways as well

<raphael> Example 3: a video segment can be a frame, a shot, a program

<raphael> Example 4: loose meaning of decomposition: not part/whole semantics

<raphael> Approaches:

<raphael> Hunter: initially in DAML+OIL and then RDFS and OWL

<raphael> ... link to the ABC core model

<raphael> ... decomposition is captured using subclass constructs

<raphael> raphel: actually, the result of the decomposition, the segment !!!

<raphael> no formal axioms: an image cannot be decomposed temporally

<raphael> SMARTWeb approach: same problem, no such axioms

<raphael> Tsinaraki: translation of MPEG-7 MDS into OWL

<raphael> Example whith the variations between Tsinaraki and Hunter ontologies

<raphael> Possible solutions: 1/ mappings between the various ontologies ; 2/ build a core ontology ? ; 3/ formalize the profiles

<raphael> Scribenick: raphael

Chrisa presents the DS-MIRF framework

Transformations from OWL/RDF to MPEG-7/21 in both ways

use of a reconstruction ontology: introducing the ordering of the elements, the default values

the solution of the use case could be based on Chrisa approach

<ocelma> +1 too!

Sophia and Chrisa will join this use case

PB: nobody knows a use of some MPEG-7 ontologies ...

Raphael: why using these two ontologies ?

Vassilis: we do not focus on these two ontologies but rather on the mechanism for bridging multimedia ontologies

Michael: what this use case is about ?

Vassilis: there is various MPEG-7 ontologies

Oscar: propose to add a new document focusing on MPEG-7
... taking the part in the vocabularies deliverable

Raphael: why not making a new document ?

<ocelma> that's what I meant! :-)

<mhausenblas> +1 , volunteer for creating it ...

<ocelma> +1 to contribute it (with Roberto Garcia, I hope!), once its created.

+1 to contribute too

Yannis: establish mappings between this multimedia ontologies and the existing vocabularies (FOAF, DC, ...)
... would be more lessons learned that can be further used rather than an example of a use case

Proposal: to take off MPEG-7 from the use case list and make a new standalone document

<mhausenblas> +1

this document will focus on MPEG-7, the various ontologies, and their interoperability problems

<scribe> ACTION: Vassilis, Raphael, Oscar, Michael, Stamatia to set up a new Deliverable [recorded in http://www.w3.org/2006/12/08-mmsem-minutes.html#action05]

Thierry and Paul and Massimo at DFKI will join this deliverable

4. Photo Use Case

Susanne Boll presents the use case

Photos: today (pixel-based + additional information)

tomorrow: capture and derive semantics annotation

What we want to know about photos: who, when, where, how long, weather, place, location, surrounding, company, quality, what I did, moud

Is tagging enough ?

there is not unique reasonable description scheme for photos (EXIF, photoRDF, IPTC)

photo undergo cross-platform distribution search

need for low-level features

Yiannis: MPEG-A about photo collections

<mhausenblas> http://www.chiariglione.org/mpeg/technologies/mpa-pp/index.htm

need for context and content

<ocelma> (similar as Music UC)

Existing standards: EXIF, IPTC, PhotoRDF, FOAF, DC, MPEG-7

scribe:

Interoperability problems:

scribe: metadata preservation and enrichment
... centralized vs decentralized storage
... lacking lean but comprehensive standard

What could be the solution ?

Face recognition is really important, as well as Event (using potentially contextual information)

Propose to have a standard way to include in the photo a link towards the description

scribe: so that photo application system can use the metadata to do ... cool stuff :-)

What kind of metadata ?

Focus on EXIF, a standard for visual features, a vocabulary for the content ...

Raphael: will you use RDF at all ?

Susanne: I'm not against, but not big fan !
... my concern is about the size of the description

Erik will join this use case

<scribe> ACTION: Susanne to look again in the various systems and standards and identify 3-4 standards that together could be the solution [recorded in http://www.w3.org/2006/12/08-mmsem-minutes.html#action06]

5. Multimedia Authoring Use Case

SMIL, SVG, Flash, SMIL BLP, SVG Tiny

Example of a SMIL document

Problem: you use metadata in the select and assemble processes, but there is a semantics barriers
... as soon as you transform it in SMIL or in Flash ... you loose the metadata !

What is lost: structural information is hidden in the format of the presentation, emergent semantics when authoring

What is needed: making discourse and structural information explicit for search and access to documents and fragments

Existing standards: LOM (Learning Objects metadata) and SMIL metadata module

Multimedia documents hide almost everything ... no description for search

scribe: an annotation problem, as well as an interoperability

Proposal: is this use case a clear open issue ?
... no solutions seem exist
... we all clearly see the problems
... but not the solutions
... susanne will continue to edit her thoughts on the wiki

<scribe> ACTION: Erik to remind Susanne about what is this standard your group has developped ! [recorded in http://www.w3.org/2006/12/08-mmsem-minutes.html#action07]

6. Semantic Media Retrieval Use Case

Tobias will join the previous use case (Multimedia Authoring)

Ioannis Pratikakis

(From the Analysis Perspective)

Deal with semantic interoperability for automatic media interpretation / understanding

linking of low-level features to high level semantics

Example 1: lack of semantics in low-level descriptors (single media)

Example 2: interchange of semantics between media (multiple media)

This use case related to interchange of low-level descriptors for semantic interpretation

scribe: exchanging classifiers is more interesting that exchanging descriptors
... MPEG-7 is useful for content-based retrieval
... but not for any semantic access

Interoperability problems

A user might want to exchange class descriptions between an imaging device, a set-top box, etc ... the process should be exchanged

Solution:

exchange classifier parameters, and class descriptions

sketch of solution: multimedia core ontology, media descriptor ontology, classifier ontology, domain ontology

Raphael: sounds really close to the algorithm use case

Example contained in the use cases

Vassilis: sounds like more a fusion problem that an interoperability problems
... will be discussed more tomorrow with the algorithm use case
... seems to have a lot of overlap

That close the meeting

Meeting starts tomorrow at 9:30 Greek time !!!!

See you tomorrow

- DRAFT -

MMSEM Athens F2F Meeting, December 8-9 2006

8 Dec 2006

Attendees

Contents