author: Joost Geurts date: 6-6-2005 one day workshop on Multimedia and the Semantic Web: http://www.acemedia.org/ESWC2005_MSW/ 2nd Annual European Semantic Web Conference: http://www.eswc2005.org/ Sunday 29 may 2005 (workshop) Overall opinion: I think the attendees of the workshop can be roughly divided into three groups: logicians, feature detection people and application oriented people (including us). Although these traditionally are quite separate groups the atmosphere was not hostile at all and people discussed quite openly. In some sense people were exploring the area to see what others were doing. An interesting discussion for example was whether multimedia data can be part of an ontology to define a concept. From a logicians perspective it can not because the semantics need to be described in logic so you always need to create a model. From a user perspective however defining/grounding concepts might work better with other media. (e.g. concept colour 'blue'). My talk was scheduled right after Monica's keynote, so I was the first presenting a paper. People were fresh, payed attention and had some good questions. Which surprised me was that they liked to know a little more about annotation schema's which I had not really discussed during the talk because I though it was too complicated. People were interested in the linking stuff and recognized the problem but didn't really have an answer. At some level my talk kind of summarized the other talks since a number of problems reoccurred. There was time for discussions, not too much though and most of the *real* work was done during coffee breaks and after the workshop. What I felt missing was somebody who could orchestrate things and define something like action points or so. After the workshop we discussed with some people the possibility of defining a mm ontology. The idea was to get three or more bigger ontologies and see whether we can find a common base. The result was that we could not find three ontologies which were open source licensed so in the end we did not make any concrete plans. Nevertheless things are starting to move so we need to keep an eye open for it. Furthermore it was good to find out that people know us and that they liked our work. Talked quite some time to Jeff Pan who works on rules for the semanticweb and he was interested in what types of rules we would use in a mm environment. Giorgos Stamou, the editor of Multimedia Content and the Semantic Web: Standards, Methods and Tools http://eu.wiley.com/WileyCDA/WileyTitle/productCd-0470857536.html joined us later and he seemed quite interested in our application as well. Conference: The talks in the conference were not earthshaking. In my opinion they made a scheduling mistake by having a complete day reserved for the industry forum while at the other two days there were three parallel sessions which sometimes caused that you missed a talk which you wanted to see. There was a kind of interesting talk about temporal logic in RDF. So what happens for example if subclass relationships only hold for a period of time. Furthermore Anita de Waard of Elsevier gave an interesting talk about how the web can change the authoring paradigm. I did not agree with all of her statements and we had some debate during the conference dinner but in the end I think we might be after the same thing. I kept seeing relations between film production and the acquisition of meta data during the process and her idea authoring documents not tied to the currently established authoring process, but instead write in more dynamic way where the document is accessible at every stage during the authoring process. The demo/poster session was quite good. I saw a nice demo of the simac project (http://www.semanticaudio.org) which exported low level feature detection to RDF triples and then could combine them with higher level semantics. In general the atmosphere was quite relaxed (hey we were on Crete) it had this air of a small conference were you can talk to everybody and join any conversation which I liked. I have the proceedings, so you are welcome to have a look. This are the raw notes I took there are some names stefano, katya to indicate points of interest. ------------------- RAW notes ----------------- sun 29 may (workshop) book: MM content and the semweb, methods, standards and tools http://eu.wiley.com/WileyCDA/WileyTitle/productCd-0470857536.html Guus Schreiber use cases in e-culture: q: find similar chinese chair (time/space) tech req: thesauri relating time/space concepts nlp for unstructured descriptions time/space reasoning q: find paintings of similar style annotations do not relate them -> use background knowledge issues: match artist to ulan:artist ok artist - style link: aat contains styles, ulan contains artist: there is no explicit link learn link from corpora derive it from annotations need domain dependent reasoning rules issues wrt thesauri - public availability - rdf/owl representation - leaning/specifying term/concept mapping mannage evolution q: find images with the same subject (e.g dancing) cocneptual subject descriptions general - descrfiption of generally known items '"ape eating banana" specific - descriptions of large black, ae named xxx eating a banana from congo... abstract - to difficult in practise -- show hierachy to disambiguate term mian observation: a combibation of different techniques is needed to be able to cope with the complexity of mm semantics nlp, segmentation, cbr, vis feature .. key role for user studies Sculpteur CIDOC SRM to describe semantics *****Using several ontologies for describing audio-visual documents owl dlp module for sesame Semantically enhanced tv news - christian ursu *****A mm adaption framework based on semweb technology - dietmar jannach personalize content monday 30 may ***** paula hobson, industry requiremenst for mm. applications - search, support of e-commerce, enabling automatation of transaction ***** Semantic Web, Use Cases and Challenges at EADS for Aerospace -- Anne Monceaux, EADS, France what do we need? easy delpoyment, reuse of legacy -> mediation technolgy meacnism to reason over answers: combone info, KR expertise: ontoloy based representatuib corese - semantic web server(?) inria - search - retrieve - cluster wanted: user advantages for reasoning rules **** Science Publishing and the Semantic Web, Or: Why are you reading This On Paper? - Anita de Waard, Advanced Technology Group, Elsevier Amsterdam overview of histroy of documents very few people treat digital docuemnts diferent from traditional document. digitliazitation is just a way to transfer it. book is equivalent to truth/value - copy feals cheap and has not the same value independent of the content pay for documens in terms of how valuable they are intelctually (how much tim to produce) telephone analogue docuement is not knowledge- the enities and relations within are - write document which explain relations direclty (linking) versioning is often ignored! you currently need an introduction to scope your research. This is hard becasue you ned to leave out stuff. What if the authoring paradigms links to the stuff it extends on. (what about context? different interpretations? outsiders of the field?) text processing is like food processing - mixing the content and sucking it up through a straw. need to develop methods which maintain the flavour need for a new authoring model which publishes not only the end results but produces evolutions ***** Semantic Integration in the real world Jürgen Angele, Ontoprise GmbH $1 for development, integration $5 - $9 problems: structurla hetroginity user query: give a single view on .. - aggregate dat a from differnet sources adapted to vocabulary of user - show different perspectives should an insurance policy be canceld? he is a lousy payer BUT his cousing is CEO of an important client -> do not cancel policy! single view - buisiness agility minimize impact of change, ease of maintnacnejnce , rapid implemntation of new strategies increased productivity *****"Towards a theory of formal classification" was SWebB: Semantic Web Browsing Fausto Giunchiglia - Universtity of Trento target problem: mannaging semantic hetrogenity in the large lightweight is a must target application: p2p systems classifications have been used for centeries by librariens and archivist methodologies are optimized for humans (dags, multirooted trees with facets) a formal classification (fc) is a rooted tree where each node is assigned a label in a propistional language moving fron natural langiage to logic brings advantages: nl is ambiguous logic has set theoetic semantics a label in logic is a propisitional formula conclusions: classifications are importatn classification encode xonomic knowledge in an efficient way normal formal classification are lightweight taxonomies ***** Tailored textual summaries from ontologies -aklina NLG takes as input structured data and outputs a text system ontosum read paper: adapts nlp to semantic web envirionment (used to be traditionally proprietary knowledge sources) http://www.sekt-project.com does formatting of simple message (e.g use a list when there is a enumaration) implemented in prolog ***** AquaLog: An Ontology-portable Question Answering System for the Semantic Web *Vanessa Lopez, Michele Pasin, Enrico Motta problem is dealing with terms which are not known to the ontology. uses wordent and google to interpet unknown terms ***** Lexicallly evaluating ontology triples generated automatically from textsa Peter Spyns, Marie-Laure Reimberger [[stefano]] automatically generates vocabularies/ontologies by statistically analyzing texts starlab ontology evolution of automatically generated ontolgies on ontology conssist of a double articulation: an ontology base intuitive plausible domain facts an commitment layer example: member take mesuer vocabularie should be based on the middle region of words (statistically orderder by occurence) test measurements: recall precision coverage - are the tripples retrieved representing the domain accuracy - tripples retrieved not to general but reflecting specialized terms result: the output is reliable but needs to be complemented with another source discussion merics: simple metrics for evalution automated generarted onotlgies weakness: rahter rough metric (but how precise should it be?) ***** Temporal RDF Claudio Gutierrez, Carlos Hurtado, Alejandro Vaisman graph representing temproal relations e.g. phd student is a superclass of person ''mary" during time t1..t3 two aproches two model in rdf: versioning - label element subject two change vs labeling - keeping snapshots of each state of the graph time point vs. time intervals (need both) query - atudents who take a master course between 2000-2001 issue: large collections of RDF data are not easily available ***** Semantic Annotation of Images and Videos for Multimedia Analysis Stephan Bloehdorn, Kosmas Petridis, Carsten Saathoff, Nikos Simou, Vassilis Tzouvaras, Yannis Avrithis, Siegfried Handschuh, Yiannis Kompatsiaris, Steffen Staab, Michael G. Strintzis map low level feature to high level semantics (lookS like our reqs) reqs: - asoc. vis features with concept descriptors - user friendly annoation - modularization - linking into mm. KNOWLEDGE INFRA STUCTURE DESIGN: MM STUCTURE ONTOLOGY domain ontologirs prototype approach system M-omtomat-annotizer q: how do you deal with mismatches (too general vs too specific) a: that is a problem in the current situation we are looking a fuzzy logic to resolve some of the issues ***** Towards Semantically-Interlinked Online Communities John G. Breslin, Andreas Harth, **Uldis Bojars**, Stefan Decker **** The Personal Publication Reader: Illustrating Web Data Extraction, Personalization and Reasoning for the Semantic Web Robert Baumgartner, **Nicola Henze**, Marcus Herzog peter publishes the same information about his research on different resources problem data is duplicated solution: publish info once - and reason about different publishing methods/presentation make interaltions between differen resources. e.g two papers are realted because the authors work together in a projects steps: 1) getting the datat 2) rules for a syndicated view 3) integrate in a semannticweb enabled view extract info from html pages, store in RDF Wednesday June 1 **** DRAGO: Distributed Reasoning Architecture for the Semantic Web Luciano Serafini, Andrei Tamilin **** Feta: A light-weight architecture for user oriented semantic service discovery Phillip Lord, Pinar Alper, Chris Wroe, Carole Goble **** An Argumentation Ontology for DIstributed, Loosely-controlled and evolvInG Engineering processes of oNTologies (DILIGENT) *Christoph Tempich*, H. Sofia Pinto, York Sure, Steffen Staab [[stefano]] diligent argumentation ontology available? aims at a vocabukary to control distubuted ontology creation. - dsitributed, evolving , loosly controlled, engeneering of ontologies requirements on th evocabulary: common vocabluray (existing vocabulary)n (e.g IBIS, Compendium) relevenat arguments (e.g Example vs ccounter Example) adaptive (manual vs automaed provision of arguments) support entire argumentation conceptual as well as fomralization (expresivity vs computational ?) formalism independent process awareness sekt project http://www.sekt-project.com http:/www. ceno argumentation framework ´97 ´98 **** Collaborative and Usage-driven Evolution of Personal Ontologies Peter Haase, Andreas Hotho, Lars Schmidt-Thieme, York Sure scenario: bibster ... ****Debugging and Semantic Clarification by Pinpointing Stefan Schlobach [katya, stefano] (read paper) building ontologies is difficult suporting debuging explain errors/incositenties uses owl/dl logic vocabularies experiments with sumo, cyc, soccer ... **** Web Explanations for Semantic Heterogeneity Discovery Pavel Shvaiko, Fausto Giunchiglia, Paulo Pinheiro da Silva, Deborah McGuinness semantic matching **** Multiple vehicles for a semantic navigation across hyper-environments Irene Celino, Emanuele Della Valle (discussed poster cocoon during poster session, cocoon has nothing to do with cocoon system) the web is not a personal space (its structure is unkown to the user) web is a graph not a tree (structure is lacking) portals provide some of this structure. but need to be adaptable navigatio model usefull to categorize links access model - usefull to cretae guided tours presentation model - used to create graphical presentations travel objects are like narrative units except that the structure can be a graph. In fact they look more like lenses. model - view - controler architecture design pattern http://seip.cefriel.it q: why does the navigaion stucture look like the ontology strucuture, can't you use the ontolgy for that a: