========================= = Oldenburg, Germany = = Visit to Susanne Boll = = 22/02/2007 = ========================= 1. OFFIS: --------- OFFIS is a non-profit research institute in computer science, non part of the university. It is a registered society. 70% of the funding are international, national and regional projects 6 R&D faculties (kind of themes): Susanne is "Multimedia and Internet Information Services" (MI) MI has 3 focus: (MI is around 16-20 persons) - Mobile Multimedia Applications (mainly Susanne, around 6-8 persons) - Technologies for eLearning - Internet in Economy, Law and Society 2. Canonical Processes: ----------------------- Lynda introduced the general model (9 canonical processes). Need for a double feedback: - all papers should follow the model (not the names of the processes but the descriptions in terms of input/output) - from the authors: is there some missing process? is there more links between the process? The processes are defined in terms of their input/output structures and the actors involved. Example: follow the table made by Alan Smeaton's group with the "SenseCam Visual Diary Lifecycle" Lynda: we are more interested in the input/output of each process rather than what's happening in each process Canonical processes: - Premeditate: what is in your head . Susanne: but this process can has an input! For example, a schema/template (or any formal instructions) that will guide the capture. . Lynda: we could add that, but only if there is at least one system that uses it! . Question: where this template/schema should be? An input of the capture process? - Create Media Asset: capture, generation or transformation . Susanne: what is a media asset? . Lynda: Definition of "media asset" is missing. To be formally defined ... . Susanne: why not naming this process "Create Process Artifact" ? since it can be basic or composite ... - Annotate: . Alan Seaton: capture => time stamped SenseCam images; processing => low-level visual features analysis . Susanne: I would like that the "media asset annotation" also be an input of the "Annotate" process . Zeljko: either we add a new link, or we give up with the difference between "PremeditateAnnotation", "CreationAnnotation", "MediaAssetAnnotation", etc ... but just have a "ProcessArtefactAnnotation" ? - Package: . Susanne: Why not making "Package" a subclass of "Create"? that comes back to what is the "Create" process? - Query: . Susanne: "query" could mean for some communities "exact match". What you want to say is "select"? . Susanne: is there a notion of ranking? ... perhaps in the "Organise" process? - Construct: really linked to premeditate . Zeljko: is close to post-production - Organise: . Lynda: the ranking (as Information Retrieval see it) is part of this process - Publish: . Susanne: is "Publish" implies to use several media assets? If I tag a photo on Flickr and I would like to send it to my mom, is it "Publish"? . YES because the photo + tag is a new document - Distribute: . Susanne: not sure I got the difference with "Publish" . Is saveAs Flash/SMIL presentations a Publish or a Distribute? Questions: - Susanne: where do you capture annotation cross-media (e.g: these 5 photos belong to a category or are related somewhere)? . Using the "composite artefact" Summary: The big clashes are: - "Create Media Asset" or "Create Process Artifact": every process "create" things (e.g.: Annotate creates Annotation, a BasicArtifact). Would we like to maintain the distinction between the unit and the composition? - Does the "Package" process really necessary? - What is exactly the distinction between "Publish" and "Distribute" process? ACTION: define the concept "media asset" more precisely, perhaps with an exhaustive list of what could and could not be media asset? ACTION: see in the general model where "Media Asset" should be renamed into "Process Artifact" ACTION: should we really need the "Package" process in the model? ACTION: Lynda to write something about the differences between "premeditate / construct message / distribute" ACTION: ask "Svetha Venkatesh" to submit a contribution to the special issue Discussion with Frank / Ansgar: - Frank: "Create Media Asset" can be renamed as "Create Process Artifact" - Frank: "Package process" can be deleted - Frank: "Query" can be renamed as "Collect" ? - Frank: "Publish": as soon as one makes a document available ... the way it is accessible is another business, the goal of "Distribute"? - Ansgar: what the differences are between "Publish" and "Distribute" - Ansgar: why there is a "Premeditate"? ACTION: Zeljko, Lynda to give answers to Ansgar to these questions by mail 3. Mutimedia Semantics XG: -------------------------- Reactions of Susanne: - It is difficult to phrase the problems even though it becomes to be clearer in my mind ACTION: Raphael to ask speakers to identify themselves before speaking Positive: . Susanne is now thinking of interoperabilty and standards when developping technologies . Susanne would like to have more physical meetings / workshops 4. WWW 2007 Panel: ------------------ Date? wait for the final decision Suggestion: do we use web technologies? People type questions in live? Setting up a Wiki? 5. Demos: --------- ACTION: Raphael to send reference about PiggyBank to Susanne (geo-localisation of buildings and display on maps) Photo use case: slides are really interesting! Susanne should describe that in the use case! Open questions: what are the formats used in each box? What are the possible formats? In the use case, alternatives should be described ... and interoperability problems shown. Context information: use of EXIF, geo-localisation, calendar? => gives examples of format (vcal for calendar?) Content information: shows examples of format used (even the proprietary ones!). Why the input/output formats of the various analysis boxes should be exchanged?