See also: IRC log
<JeffP> this will be mmsem First F2F
<JeffP> what conference is this?
<raphael> Scribe: jacco2
people on IRC: you can connect by phone too!
See phone number in the agenda
<pasquale> sorry, for now i can only irc
Chair opens the meeting and welcomes all participants
start with role call
Yioannis, athens, greece
, feature extraction
<JeffP> I cannot hear very good
Vassilis, ITI, Greece, research interest SemWeb technologies and mm annotation
Giovanni, DBin project, DBin is p2p metadata exchange
Christian, PhD student, also from Ancona
Thomas, Koblenz, also in k-space and aceMedia - mediator role
thierry, DFKI, aslo in k-space. Member of ISO group on language resources
George, CWI, PhD student, also in k-space
Lynda, CWI, k-space
Jacco, CWI, Semantic Media Interfaces
Jeff Pan, chair of XG, semantic web languages, reasoning
Officially, we now have 21 members
Vassilis of NTUA is AC Rep for k-space members
Giovanni, membership is pending via Italian W3C office
We also have Susanne Boll as an invited expert
<raphael> hi Vassilis, do you plan to join on the phone ?
<VassilisTzouv> Hello all
<JeffP> Hi Vassilis
<VassilisTzouv> Yes I will join on the phone
Vassilis and Yiannis connect by phone & IRC, have send slides which will be presented later
<jacco> Chair introduces W3C processes
<jacco> Explains about zakim and RRSagent: all members, please lookup the short explanations on the W3C website
<jacco> each member will be asked to scribe
<jacco> we will have bi-weekly teleconf, which will be announced on the web page and mailing list
<jacco> public minutes will be published
<jacco> Chair explains goal of the W3C incubators.
<raphael> Incubator Activity: http://www.w3.org/2005/Talks/1130-sb-w3c-ac-mangrpt/Overview.html#(15)
<jacco> Raphael shows slides of Steve
<jacco> Important: XGs are light-weight, and will finish within one year
<jacco> Goal is to publish technical report with advice to W3C about potential Recommendation track followup
<jacco> We are the second XG within W3C
<jacco> jeff: xg is new way to prepare a new activity, with much better infrastructure as before
<jacco> jeff: OWL was based on member submission on DAML+OIL, but it took still 2 years to get the OWL REC out, which is not that much difference
<jacco> With the XG we have a better infrastructure to do the ground work of the preparation
<jacco> So later during this F2F we should discuss if we want to try to setup a Working Group
<jacco> ... after the charter of the XG ends
<jacco> Raphael: the Content Labeling XG was the first XG, labeling content for parental guidance etc
<jacco> We plan to work closely with this XG, invite all members to follow the work of this XG
<jacco> This XG is the second XG, last week the 3rd XG started on "GEO", geo-tagging etc
<jacco> Raphael shows charter
<jacco> Explain history of the MM taskforce of the SWBPD group
<jacco> Goal 1 of this XG: show how semweb can be used to achieve mm metadata interoperability
<jacco> Non-goal: create new MM metadata standards
<jacco> Goal 2: show added value of formal semantics by describing practical applications and examples
<jacco> Goal 3: provide best practices based on the above
<jacco> Open questions: the modalities do we address, what standards, etc
<jacco> Intro of Massimo: I'm from CNR Pisa, multimedia understanding, Semantic Web and XML-based languages
<jacco> Yioannis:W3C lives too much in its own world, need to talk to other bodies more. In addition, related standards like mpeg-7 have not yet finished. So we need to communicate to these bodies
<jacco> So how will this happen?
<jacco> Raphael: I agree, best way is to have representatives from these other bodies in this XG as a member or invited expert
<jacco> But I do not know who could best represent the MPEG-7 community
<jacco> Yioannis: we could recommend a view names
<jacco> Yionannis: I think interoperability is the key issue, we would focus on automatic annotation
<JeffP> Can people say their names before they speak please?
<jacco> Vassilis (ITI): I agree with previous speaker. Not only focus on the RDF-level, also OWL, Rules or extensions.
<jacco> Modalities: we could start with images and then try to extend later
<jacco> jeff: agrees with Vassilis
<JeffP> Jeff: ... and OWL and rules are also considered in the charter explicitly
<jacco> Raphael: MMTF of the SWBPD only did images, maybe do this again is not a good idea
<jacco> Giovanni: Wants to stress the importance of real use cases, not focus on theory & formalisation side
<jacco> Giovanni: I've tried the music-oriented sites, often felt lost. We could find good use cases in this area. Content-based identification is important, once we have a URI there is a lot we can do
<jacco> What do we identify? Melody, wave signal, midi or mp3 file, etc
<jacco> Raphael: what is CDDB using? Giovanni: its based on a fingerprint of the recording, not very robust on for example other versions of the same track on a different album etc
<jacco> ACTION: Giovanni to coordinate write up of a use case on music [recorded in http://www.w3.org/2006/07/10-mmsem-minutes.html#action01]
<jacco> Christian: Agree with Giovanni, real use case in a distributed may also need to address right management
<jacco> Giovanni: not everybody want all metadatato be open and free
<jacco> Raphael: Does the group care about rights management?
<jacco> Yionnis: addressing MPEG-21 is also part of the interoperability issue. Doing music means we expand from images to audio?
<jacco> Thomas: different use cases with different modalities
<jacco> SemWeb could be a good opportunity to also show high-level metadata intergrating multiple modalities.
<jacco> For example, using flickr annotations and than connecting it using semweb technology could be a good basis for a use case
<jacco> Flickr is a good example to show how easy things can be
<jacco> Thomas: our group is interested in combining modalities, like image + audio
<jacco> Thierry: I'm not a specialist on multimedia. Analysis of lyrics or speach transcripts could be used for this use case
<jacco> Thierry: we could also look at other projects for good use cases
<jacco> A use case from, for example, the MESH project, could be on news.
<jacco> Giovanni: MUSCLE is working on music use cases
<jacco> AXMEDIS is a big player in this
<jacco> Massimo, could you say something on MUSCLE
<jacco> within the MPEG-7 standard actually using the description schemes is really difficult
<jacco> ACTION: thierry to write up limitations of MPEG-7 description schemes [recorded in http://www.w3.org/2006/07/10-mmsem-minutes.html#action02]
<jacco> George: also wants to focus on real use cases, like the tagging example mentioned by Thomas
<Massimo> in MUSCLE we are developing an infrastructure to manage multimedia metadata (images, audio, video). We started using the MPEG-7 standard.
<JeffP> shall we have a use case for tagging?
<raphael> Massimo, do you have concrete use cases written in MUSCLE ?
<raphael> About MUSIC ?
<jacco> george: blogging could be another use case, with different modalities
<Massimo> I think we should remember the use cases of the Semantic Web: eg: catalogs of personal photos,
<Massimo> systems able to recognize persons' faces
<jacco> george: DRM: I can see the concern, but it is a hard issue that will not be solved by this group
<Massimo> and then a thing we haven't still seen that is about processed needed to obtain multimedia objects
<jacco> giovannni: need to align social/human tagging with formal/machine tagging
<jacco> Raphael: was a workshop on this topic at WWW conference, this community starts to pick this up.
<lynda> jacco: images too limited, so include other modalities
<lynda> bridge gap between formal ontologies and tagging
<Massimo> about music thinking of the semweb use case the personal catalog and the catalog of groups (sorry i can't hear i don't want to break who is talking)
<jacco> ACTION: george to write up a use case around tagging [recorded in http://www.w3.org/2006/07/10-mmsem-minutes.html#action03]
<jacco> raphael: summary: focus on practical use cases like music combined with lyrics analysis and lower level feature extraction, use case on social tagging and semantic web
<jacco> Yioannis: our features are not limited to the ones defined by MPEG-7
<jacco> Giovanni: when you use additional features, there is no incentive to go through all the pain of a standardized serialization, becuase there is no aother tool which can read it anyway
<jacco> Giovanni: this is related to the fact that MPEG-7 semantics is described in informal English only, not machine readable version.
<jacco> SemWeb approach could help here
<giovannit> jacco: well not really english only , there is structuring among the decriptors, but the meaning of each one is then in english
<jacco> Yioannis:mpeg-7 is discussing support transformation matrices to transform labels
<raphael> Nikolaos_Simou is Vassilis
<jacco> Vasilis: want to raise the question if we are interested in ontologicalization of low level features?
<VassilisTzouv> Nikolaos_Simou is Vassilis
<jacco> Raphael: for what do you need a semantic interpretation of the low level descriptors?
<jacco> yioannis: we need both numeric and symbolic features in various levels
<jacco> ACTION: Yioanis to write up use case showing the need for semantic interpretation of low level features [recorded in http://www.w3.org/2006/07/10-mmsem-minutes.html#action04]
<jacco> yioannis: in the "fused" model all modalities are used to improve the automatic annotation
<jacco> ACTION: Raphael to write up use case in the news domain [recorded in http://www.w3.org/2006/07/10-mmsem-minutes.html#action05]
<jacco> IPTC standardizes NewsML, now doing version 2 of NewsML
<jacco> CWI works with IPTC in the context of K-Space
<jacco> NewsML will likely be in the spirit of RDF, but based on an XML syntax
<jacco> A representative of IPTC is expected to join the XG in the near future
<jacco> Raphael: there are different sources of metadata: from the camera, from the photographer's (manual) annotation tool, news agency will add more metadata, etc
<Massimo> (trying not to break the conversation..) i think we have to remember two more things about use cases: (1) the "Image annotation on the Semantic Web" note (http://www.w3.org/2001/sw/BestPractices/MM/image_annotation_galway.html), Jacco is one of the editors and (2) Algorithms (processed) to obtain multimedia objects
<jacco> URL of last version of the note is at http://www.w3.org/2001/sw/BestPractices/MM/image_annotation.html
<Massimo> (...processeS /)
<jacco> Giovanni: we could think about form interfaces generated by local event ontologies
<jacco> massimo, what do you mean with 'obtain multimedia objects'? Object tracking & recognition?
<Massimo> i think there could be systems able to reconstruct multimedia objects
<jacco> Raphael: should we provide some demo web service for mm annotation?
<Massimo> a set of steps to re-obtain an object
<jacco> Giovanni:, yes, if we can pinpoint to a basic feature like identification, that crosses all use cases
<Massimo> thinking of images for example if an image is obtained after having applied some algorithms then we can repeat the same steps to do it again
<jacco> Maybe there will be an XG on identification as a followup of the Workshop at WWW2006 on this topic
<jacco> Raphael: Repeats things said by Massimo on IRC.
<jacco> Notes that the "Image annotation on the Semantic Web" will be discussed by Jacco in the afternoon
<jacco> Vasilis: I think massimo refers to the process of analyzing the content automatically
<Massimo> to analyze and to reproduce the content
<jacco> For example: To recognize this feature, you should apply this process
<jacco> Massimo, can you explain about "reproduce the content"
<jacco> Is this a topic in Muscle?
<Massimo> yes in Muscle and in the ace Media
<Massimo> (the call for a common multimedia ontology(
<jacco> Thierry: Raphael, what is the role of low level features in your NewsML scenario?
<jacco> Raphael: Currently none...
<jacco> Massimo, could you write up something about the need for being able to reproducing the content? Could I give you an action point?
<Massimo> ok, i'm writing it give me some minutes
<jacco> there is no need to hurry :-)
<JeffP> Massimo, for the record, could you send to the public list please?
<jacco> ACTION: Massimo to write up use case on analysis and reproducing content [recorded in http://www.w3.org/2006/07/10-mmsem-minutes.html#action06]
<jacco> Raphael: Mentioned that this XG will focus on metadata interoperability, where k-space is doing work on combining low level features
<jacco> Vassilis: are we considering cross modality interoperability?
<jacco> Raphael: yes, we discussed this in the beginning of the meeting
<jacco> Vassilis: I'm not sure this is in scope for the 1st deliverable?
<jacco> Raphael: We have not decided about what documents and deliverables we will produce
<jacco> Vassilis: if it still on the metadata level, it is ok
<jacco> Raphael: summary of morning session: people have shown interest on practical use cases on different modalities and domains, 6 action points to write up use cases
<jacco> Raphael: use cases will also look at the way RDF, OWL, rules etc will be used to solve the use case problem
<jacco> Now first lunch, after lunch we well have time for some demonstrations, after coffee break more discussion about the deliverables of the XG
<jacco> Lunch break!
<raphael> ??P0 is meeting_room
<giovannit> jeff: fine with me to start again
<jacco> scribe: Giovanni
<giovannit> (practicing the scribing..)
<jacco> scribenick: giovannit
i'll attempt my best (-:
raphael: 10 minute per person presentation
<JeffP> Are the slides available online?
<jacco> jeff, we will make the slides available asap
<jacco> Jacco show demo at http://e-culture.multimedian.nl/demo/search
ecoluter applications: converted metadata of a few museums into rdf and seen how far we get without doing heavy weight alignment
<raphael> All the slides shwon will be at: http://homepages.cwi.nl/~troncy/tmp/W3C/
jaccos'd demo shows a full text search which works recursively on the rdf graph
basically resources are collected starting from those which have a literal attached matching the query, a number of euristics and settings enable meaningful rating and limiting of the end result
the whole structure works also fine when generic dataset is loaded, providing an ontology based navigations of the instances and visualization of related metadata. Cool ajax makes it apparently very responsive
the work is powered by swi prolog. Jacco says they tried doing this in sparql but the expressivity is really not sufficient for the needs of the algorithms used here.
Ioannis presenting now
presentation introduces ellogon http://www.ellogon.org a general purpose language engineering platform
CIL: a computational intelligence laboratory, applications are shown in the area of OCR , shape based retrieval, human tracking
now presenting the BOEMIE project
low level methodologies are improoving but high level annotation extraction is still difficult, the use of ontologies to drive the information extraction process seem very interesting
<jacco> BOEMIE is, and other slides will be, published at http://www.w3.org/2005/Incubator/mmsem/meetings/f2f-amsterdam/
Among the aspects involved in this project are semantically annotated maps, evolving ontologies and fusion of different aspects of multimedia metadata
a use case/query is presented merging all the aspects covered by the project.
raphael: there are other EU proejcts which have similar goals ,what ar ethe differences?
ioannis: BOEMIE is concerned with multimodal metadata , to the best of my knowledge there is not much overlap with the other projects
question: at which level is the fusion actually going to happen
ioannis: the fusion is going to
be supported by reasoning. Mid level concepts will be
constructed and reasoning mechanism will take the decisions
about the high level concepts
... another novel concept is that the semantic extraction is knowledge driven, using an ontology
toward a common multimedia ontology framework: initiative started a few months ago , a framework for collecting existing metadata knowledge reppresentation , an approach for harmonization..
In the first phase of this initiative, a number of requirements were collected
reasoning support needs rules and DL based methoodologies but also uncertainty and fuzzyness
modularization is agreed to be an important feature. Smaller ontologies which are reusable, specific and interoperable
there is the need of a upper ontology "merging" or "harmonization". In general some practical way for interoperability since people are using different ones.
a semi automatic image annotation tool is shown. Automatic image segmentation helps a human operator to associate ontology labels (shown as a tree)
<Massimo> about the possible use case on algorithms..
<Massimo> Currently we have sets of tools to extract features from multimedia objects:
<Massimo> a set of rules can be defined able to establish which tools could be
<Massimo> combined and how (e.g. put in a pipeline).
<Massimo> Considering for example image understanding problems (e.g. segmentation,
<Massimo> analysis, ...),
<Massimo> a thesaurus containing related concepts and algorithm could be thought of.
<Massimo> It would important to represent features on how metadata could be
<Massimo> extracted so that it could be used at a higher level
<Massimo> for images and multimedia information handling in general.
<Massimo> Another important feature would be to take trace of how a specific
<Massimo> result has been obtained starting from a particular input.
<Massimo> A specific case of image elaboration:
<Massimo> it is necessary to apply to the image a series of algorithms to
<Massimo> calculate a property (region shapes, color distribution)
<Massimo> The chain of algorithms defines a procedure that resolve the given problem.
<Massimo> It would be interesting to 'store' the procedure - as an ordered set of
<Massimo> algorithms - to add 'knowledge'
<Massimo> to a system for interpretation and analysis.
<Massimo> Once stored this procedure the knowledge increases and another problem
<Massimo> could be resolved applying the procedures already inserted.
<Massimo> Finally, this knowledge could be codified using an ontology that merges
<Massimo> procedures and problems.
<jacco> Giovanni: what about a formal cooperation between this XG and the requirements work started by aceMedia
<jacco> Raphael: yes, we will. Yiannis will become a member.
<jacco> Vassilis of NTUA presenting slides at http://www.w3.org/2005/Incubator/mmsem/meetings/f2f-amsterdam/IVML.profile.2006.06.ppt via the telephone
<jacco> Yannis presenting slides at http://www.w3.org/2005/Incubator/mmsem/meetings/f2f-amsterdam/IVML.background.2006.06.ppt
Yannis speaking (NTUA) analisys of image and video content for automatic annotaion. A tool is shown for automatic segmentation, and manual region merging
<jacco> me: hi Chris
<VassilisTzouv> Hi chris
<chalaschek> Hi guys...time difference is quite difficult for me :(
<JeffP> Hi Chris
<raphael> hi Chris, welcome and thx for joining
<VassilisTzouv> what time is now in maryland?
<chalaschek> Are you guys on break now?
<jacco> Show coffee break, start again in 5 minutes with Thomas presentation and DBin
<jacco> Thomas presenting http://www.w3.org/2005/Incubator/mmsem/meetings/f2f-amsterdam/iswebAtMMSEM.pdf
<jacco> giovanni presenting slides, will put online later
<jacco> slides are on distributed man/macine cooperative mm annotation
jacco gives an over view of the semantic web best practices and deployment working group
<raphael> ACTION: ALL give comments on the "Image annotation on the Semantic Web" Working Draft [recorded in http://www.w3.org/2006/07/10-mmsem-minutes.html#action07]
<raphael> The deadline for the comments are 24/07/2006 !
<raphael> schedule: finish this draft and get it published as a Working Group Note by SWBPD in the summer
<raphael> give your comments now or never :-)
<raphael> Maintain the resources pages: Overview, Tools, Vocabularies
<raphael> In charge of this XG !
<jacco> raphael mentions DOAP as an example of things we could do
<jacco> Giovanni: what about a wiki?
<raphael> ACTION: raphael to see if the W3C wiki can be used to maintain these pages [recorded in http://www.w3.org/2006/07/10-mmsem-minutes.html#action08]
<raphael> Giovanni volunteer these pages
<raphael> Jacco: this group should explore what is there and what could be improved
<raphael> ... W3C is concerned about what remains to be done in the multimedia domain
<raphael> ... needs for more industrial involvent
jacco says w3c is somehow puzzled about multimedia, traditionally being an ISO o MPEG-7 topic, but the topic of multimedia semantics is so clearly connected with the SW activity
so w3c should probably do even more
<jacco> giovanni: metric of what the advantage of using the semantic web is might also involve just lines of code, not only performance etc
<jacco> Raphael: maybe interoperability deliverable should also be centered around clear use cases, not around the standards as it is now
<jacco> jeff: are we talking about the deliverables in the charter?
<jacco> Raphael: yes
<jacco> Raphael: I'm not sure yet how many deliverables we are going to produce and it what order
<jacco> Raphael: The proposal includes to broaden the scope to other modalities, but focus on use cases
<jacco> Vassilis: I agree to start from use cases, they do the same in RIF
<jacco> giovanni: list of things we can do (semi) automatically, with some rating on how reliable this is
<Massimo> i agree to Vassilis starting from use cases
<jacco> so we know when to connect for example to social tagging for things we cannot do automatic
<jacco> Ioannis: I think it is better to start from the use cases
<jacco> Giovanni: but the two are very similar
<jacco> For example, if you have a music scenario, with a request for "similar" music, what do you mean by similar and what types of similarity can you do automatically?
<jacco> Ioannis: but is a very broad question, use cases might focus discussion
<jacco> Vassilis: what is the focus? interoperability or the reliability of the extracted metadata?
<jacco> Ioannis: stick to the existing use cases of the image annotation draft
<jacco> Jacco: I would like to stick to the use cases discussed in the morning
<jacco> Raphael: we will discuss deadlines on the action points tomorrow
<jacco> for use cases discussed in the morning, see action points at http://www.w3.org/2006/07/10-mmsem-minutes.html#ActionSummary
<JeffP> Restaurant-CafÃ© In de Waag, Nieuwmarkt 4, 1012 CR Amsterdam
<chalaschek> enjoy dinner everyone
<Massimo> enjoy dinner and Amsterdam everyone !!! "see" you tomorrow
This is scribe.perl Revision: 1.127 of Date: 2005/08/16 15:12:03 Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/ Guessing input format: RRSAgent_Text_Format (score 1.00) Succeeded: s/Christioan/Christian/ Succeeded: s/Pann/Pan/ Succeeded: s/Suzanne/Susanne/ Succeeded: s/Wnats/Wants/ Succeeded: s/suse/use/ Succeeded: s/agood/a good/ Succeeded: s/on/in/ Succeeded: s/Noe/Now/ Succeeded: s/musiums/museums/ Succeeded: s/alightment/alignment/ Succeeded: s/projects/project/ Succeeded: s/vassalis/vassilis/ Succeeded: s/iannis/Yannis/ Succeeded: s/Yiannis/Yannis/ Succeeded: s/involvmen/involvent/ Succeeded: s/tradeitionally/traditionally/ Found Scribe: jacco2 Inferring ScribeNick: jacco2 WARNING: No scribe lines found matching previous ScribeNick pattern: <giovannit> ... Found Scribe: Giovanni Found ScribeNick: giovannit Scribes: jacco2, Giovanni ScribeNicks: giovannit, jacco2 WARNING: Replacing list of attendees. Old list: Jeff_Pan meeting_room Yannis massimo Vassilis New list: Jeff_Pan meeting_room Vassilis Chalaschek JeffP Default Present: Jeff_Pan, meeting_room, Vassilis, Chalaschek, JeffP Present: Jeff_Pan meeting_room Vassilis Chalaschek JeffP Agenda: http://lists.w3.org/Archives/Public/public-xg-mmsem/2006Jul/0009.html Got date from IRC log name: 10 Jul 2006 Guessing minutes URL: http://www.w3.org/2006/07/10-mmsem-minutes.html People with action items: all george giovanni massimo raphael thierry yioanis WARNING: Input appears to use implicit continuation lines. You may need the "-implicitContinuations" option.[End of scribe.perl diagnostic output]