WEB SEMÀNTICA

European Research Consortium for Informatics and Mathematicswww.ercim.org

Number 51 October 2002

S p e c i a l :

S e m a n t i c We b

JOINT ERCIM ACTIONS

4 The FNR becomes ERCIM’sLuxembourg Memberby Raymond Bausch, FNR

5 Framework VI ERCIM’s Expressions of Interest

5 Euro-Legalby Heather Weaver, CLRC

SPECIAL THEME

7 A Few Words about the Semantic Weband its Development in the ERCIM Institutesby Jérôme Euzenat, INRIA

9 The Semantic Web lifts offby Tim Berners-Lee and Eric Miller, W3C

11 Semantic Web Technologiesin Europe’s IST Programme 1998 - 2002by Hans-Georg Stork, European Commission

13 Activities of INTAP Semantic Web Committee in Japanby Tomihiko Kojima, INTAP, Japan

14 OntoWeb: The Thematic Network for the Semantic Webby Ying Ding and Dieter Fensel, Vrije UniversiteitAmsterdam, The Netherlands

14 Semantic Web Advanced Development in Europeby Brian Matthews, Michael Wilson, CLRC;and Dan Brickley, W3C

16 Calendars, Schedules and the Semantic Webby Terry R. Payne, University of Southampton and Libby Miller University of Bristol

18 The ICS-FORTH RDFSuite: High-level Scalable Toolsfor the Semantic Webby Sofia Alexaki, Nikos Athanasis, Vassilis Christophides,Greg Karvounarakis, Aimilia Maganaraki, DimitrisPlexousakis and Karsten Tolle, FORTH

19 Corporate Semantic Websby Rose Dieng-Kuntz, INRIA

21 Joint Research Centre of the ECto use Semantic Web for New Privacy Agentby Giles Hogben and Marc Wilikens,Joint Research Center, Italy

23 Exmo: Assessing Semantic Properties of KnowledgeTransformationsby Jérôme Euzenat, INRIA

24 CORES — A Forum on Shared Metadata Vocabulariesby Tom Baker, Fraunhofer-Gesellschaft

25 Towards Syntax-Independent B2Bby Birgit Hofreiter, Christian Huemerand Werner Winiwarter, University of Vienna

26 Extended Faceted Taxonomies for Web Catalogsby Yannis Tzitzikas, Nicolas Spyratos, PanosConstantopoulos and Anastasia Analyti, FORTH

28 Migrating Thesauri to the Semantic Webby Michael Wilson and Brian Matthews, CLRC

29 Semantic Characterisation of Links and Documentsby Silvia Martelli and Oreste Signore, ISTI-CNR

Contents

Next issue:January 2003

Special theme:Embedded Systems

Cover: Rembrandt's self-portrait as theApostle Paul. Photograph courtesy of the Rijksmuseum, Amsterdam.

Read about the relation betweenRembrandt and the Semantic Web inLynda Hartman’s article ‘HypermediaPresentation Generation on theSemantic Web’ on page 36.

3ERCIM News No. 51, October 2002

31 An Ontology-Oriented Solutionfor Knowledge-Intensive Organizationsby Emanuele Della Valle and Maurizio Brioschi,CEFRIEL-Politecnico di Milano, Italy

32 Combining XML and Description Logicsfor Describing and Querying Documentsby Rim Alhulou and Amedeo Napoli, LORIA, France

33 Harmonise: An Ontology-Based Approachfor Semantic Interoperabilityby Michele Missikoff, IASI-CNR

34 Reducing the Complexity of InformationRetrieval by Weighted Queryingby Peter Vojtáš and Jan Vinafi,P. J. Šafárik University, Košice (SRCIM )

36 Hypermedia Presentation Generationon the Semantic Webby Lynda Hardman, CWI

37 Profile-Based Adaptability in the Semantic Webby Santtu Toivonen, VTT

38 Integrating Multimedia Componentsinto a Semantic Webby Benjamin Jung, Fraunhofer-Gesellschaft, and Lyndon JB Nixon, TCD

39 Facilitating Autonomous Active Learning in Web-Based Course Environmentsby Claus Pahl, Dublin City University, Ireland

41 Concepts on Enriching, Understandingand Retrieving the Semantics on the Webby Arne Sølvberg, Sari Hakkarainen, Terje Brasethvik,Xiaomeng Su, Mihhail Matskinand Darijus Strasunskas, NTNU

42 Negotiating on Your Own Terms!by Declan O’Sullivan and Vincent Wade, TCD

R&D AND TECHNOLOGY TRANSFER

45 Visualisation Middlewarefor E-Science: A New Projectin the UK E-Science Programmeby Julian Gallop, CLRC

46 Dealing with Massive VolumetricVisualisation: Progressive Algorithmsand Data Structuresby Rita Borgo, Pisa University andISTI-CNR; Valerio Pascucci, LawrenceLivermore National Lab, USA;and Roberto Scopigno, ISTI-CNR

48 RADIUS: Looking for Robots’ Helpin Computer Science Researchand Educationby Jozef Kelemen and Aleš Kubík,Silesian University, Opava, Czech Republic

49 Consensus Creation on UniversalAccess in Health Telematicsby Constantine Stephanidisand Demosthenes Akoumianakis, ICS-FORTH

50 A Body-Monitoring Systemwith EEG and EOG Sensorsby Mária Bieliková, Slovak University of Technology / SRCIM

52 WebRemUSINE: A Tool for UsabilityEvaluation of Web Applicationsby Laila Paganelli and Fabio Paternò,ISTI- CNR

53 Digital Factories, Production Networksby László Monostori, Géza Haidegger,József Váncza and Zsolt János Viharos,SZTAKI

54 GenoStar: A Bioinformatics Platformfor Exploratory Genomicsby François Rechenmann, INRIA

EVENTS

56 SEKE’02 Conference Reportby Filomena Ferrucci, University of Salerno, Italy

57 The Trading Agent Competition –TAC 2002by Joakim Eriksson and Sverker Janson, SICS

59 CONCUR 2002 Conference Reportby Luboš Brim, CRCIM

60 CEOI 2002 - Central EuropeanOlympiad in Informatics 2002by Gabriela Andrejková and Jan Vinafi, P. J. Safárik University, Kosice / SRCIM

60 Announcements

63 IN BRIEF

Tim Berners-Lee and Eric Millercontributed the article ‘The Semantic Web lifts off’ on page 9.

Timothy J. Berners-Lee holds the 3ComFounders chair at the Laboratory forComputer Science at MIT, and directsthe World Wide Web Consortium, anopen forum of companies andorganizations with the mission to leadthe Web to its full potential. He is theauthor of Weaving the Web.

Eric Miller is the Activity Lead for theW3C World Wide Web Consortium’sSemantic Web Initiative.

Pho

to: L

eFev

re C

omm

unic

atio

ns

4 ERCIM News No. 51, October 2002

JOINT ERCIM ACTIONS

The National Research Fund is apublic establishment with scien-tific, financial and administrativeautonomy, set up by the Law of 31May 1999 in order to further stim-ulate research activities inLuxembourg. To fulfil thismission, the Fund develops multi-year research programs and assurestheir implementation by allocationof the financial means put at itsdisposal.

Five national multi-year priorityprograms are currently ongoing,among them ‘Security and effi-ciency of new practices in e-commerce for all socio-economic actors(SE-COM)’, which has a budget of 7 500000 Euro for five years. In addition, theNational Research Fund subsidisesaccompanying measures to strengthenthe general framework of scientificresearch in Luxembourg, eg subsidisingthe organisation of scientific conferencesand the mobility of researchers.

In ERCIM, the National Research Fundrepresents researchers in Luxembourgworking in informatics and mathematicsat the following institutions:

The CRP-Santé is a public researchcentre in Luxembourg whose mainresearch and development activities liein the fields of public health, health care,and biomedicine (‘CRP’ stands forCentre de Recherche Public). The CRP-Santé aims to promote scientificprogress, technology transfer and inno-vation within the framework of interna-tional and European research and devel-opment programs.

The CRP-Gabriel Lippmann is one ofLuxembourg’s leading players in thefields of scientific research and tech-nology transfer. The centre focuses itsactivities on some of the most crucialtopics of interest of the 21st century,including nanotechnology, novel mate-

rials, sustainable management of naturalresources, biotechnology, and last but notleast technologies of the informationsociety and their relevant legal frame-works. As a public research centre, itmakes a point of keeping its commitmentin line with the economic usefulness ofits scientific achievements, howeversignificant they may be.

The CRP-Henri Tudor, founded in 1987,is an integrated centre of applied researchand technological innovation. Its mainactivities range from research to tech-nology transfer including support forentrepreneurial activities and an incu-bator for start-ups. 170 employees areworking at the centre, of whom 140 areengineers active in the fields of informa-tion and communication technology, andindustrial and environmental technology.

The Centre Universitaire is a publicinstitute in Luxembourg offering univer-sity studies. It consists of a number ofdepartments: law and economicsciences, business administration andcomputer science, literature and humani-ties, science, professional training forlawyers, and professional training forteachers.

The Institut Supérieur de Technologie(IST) is a public institute for higher tech-

nical education composed of four depart-ments: electrical, mechanical, and civilengineering, and computer science.Besides teaching, the IST also activelyperforms fundamental and appliedresearch. Several collaborations existwith different national and internationalinstitutions, both in teaching andresearch. Close relationships are alsomaintained not only with numerousEuropean universities, but also withindustry in Luxembourg and especiallywith the CRP Henri Tudor for thepurpose of technology transfer.

The implementation of a new Universityof Luxembourg with a major focus onresearch activities is currently in prepa-ration and a new law will restructure theinstitutions of higher education inLuxembourg.

Links: FNL: http://www.fnr.lu/CRP-Santé: http://www.crp-sante.lu/CRP-Gabriel Lippmann: http:// www.crpgl.lu/The CRP-Henri Tudor: http://www. tudor.lu/Centre Universitaire: http:// www.cu.lu/Institut Supérieur de Technologie:http://www.ist.lu/

Please contact: Raymond Bausch Fonds National de la RechercheTel: +352 26 1925 31E-mail: [email protected]

by Raymond Bausch

FNR, the Luxembourg National Research Fund, (Fonds National de la Recherche)joined ERCIM in July 2002.

The FNR becomes ERCIM’s Luxembourg Member

ERCIM News No. 51, October 2002 5

JOINT ERCIM ACTIONS

EURO-LEGALNews about legal information relating toInformation Technology from Europeandirectives, and pan-European legal require-ments and regulations.

Metatags that cause Trademark Infringement The UK has seen a recent landmark case thatcontains the first judicial analysis of trademarkinfringement where the use of the mark wasmade via metatags. Website owners rely onmetatags to ensure that search engines bringvisitors to their sites. The search engines usespider software to create indexes of cataloguedwebsites by reviewing the sites’ metatags forkeywords. In some cases these metatags couldinclude a trademark and website owners can usethis as a method of redirecting internet users totheir site. In order to establish trademarkinfringement, it is necessary to show that theinfringing name or mark has been used ‘in thecourse of trade’. This landmark case rules thatdespite the fact that these metatags are invisibleto the website visitor they are visible when trans-lated into a search result on the user’s browser.This case provides further protection to regis-tered trademark owners from other websitesusing their brands to attract visitors to their site.

Internet Sales — New Rules for EU Consumers The Brussels Regulation, which came into forceon 1 March 2002, introduced new rules for deter-mining where claims arising from civil andcommercial disputes involving parties fromdifferent countries can be heard within the EU.The rules apply to claims made by consumersarising out of sales of goods or services over theInternet.

Court proceedings must be brought within theMember State where the consumer lives. Thisright of jurisdiction cannot be waived even if theterms of the contract imply something else. Untilnow the contract must have been concludedwhile the consumer was in his or her MemberState. The new regulation, however, introducesan additional jurisdiction that deals specificallywith cross-boarder sales to consumers over theinternet, where the trader “directs such activitiesto that Member State or to several MemberStates including that Member State”. In otherwords, if a consumer from the UK purchasesgoods on the internet whilst on holiday in anotherMember State (eg France) he or she can bring aclaim against the trader in France providing thetrader "directed his activities" to the home Stateof the consumer (in this case, the UK).

Although this has not yet been tested in thecourts it is likely to impact more heavily ontraders who offer their goods or services in anumber of languages and accept payment indifferent currencies, including the Euro. If traderswish to avoid claims in the future, they may haveto consider accepting orders only fromconsumers residing in their own Member Stateand in the currency of that Member State. This islikely to impact on the value of trading on theinternet. It will be interesting to see how case lawdevelops in this area.

by Heather Weaver, CLRCTel: +44 1 235 446151E-mail: [email protected]

FRAMEWORK VIERCIM submitted ten Expressions of Interest informing the Commission of itsintention to propose a series of Networks of Exellence within Framework VI.Interested scientists should contact the ERCIM office at [email protected].

Anaerobic Digestion: MultidisciplinaryNetworking for Better Understandingand Optimal Monitoring — understAnDObjectives: This proposal aims at estab-lishing Europe as the scientific leader inAD processes through the sharing ofexperiences and efficient networkingbetween European leaders in multidisci-plinary fields. The main goal will be tolower the fragmentation of Europeanresearch and to increase the scientificunderstanding of the AD process, leadingto a higher competitiveness of Europeanindustry in this field.Co-ordinator: Jean-Philippe Steyer, INRA.

Applications of ComputationalMathematics and Statistics in Scienceand TechnologyObjectives: The aspiration of the networkis to link together mathematicians workingin national laboratories, universities, andindustrial R&D departments.Co-ordinator: Mario Arioli, CLCR.

Cross-Language Evaluation Forum — CLEFObjectives: The network for multilingualcross-language system evaluationintends to stimulate and assist the devel-opment of mono- and cross-languageinformation retrieval systems workingwith European languages and languagesof importance to Europe.Co-ordinator: Carol Peters, CNR.

NoE for Digital Libraries — DELOSObjectives: Digital l ibraries shouldenable access to all human knowledgeany time and anywhere, in a friendly,multi-modal, effective way, overcomingbarriers of distance, language, andculture. The network will define a jointprogram coordinating the activities ofEuropean research teams with the aim ofdeveloping next generation digital librarytechnologies, thus making feasible theimplementation of this grand vision.Co-ordinator: Costantino Thanos, CNR

European Excellence in HealthInformation Technology — E2HealthObjectives: This proposal intends to rein-force European expertise in the develop-ment of intelligent, mobile and securesystems and services aimed at providingpatients with personalized healthcare andinformation services, and at stimulatinghealth promotion and disease prevention inthe general population. The network willinclude experts in key fields in order tostrengthen and develop community scien-tific and technological excellence. Thenetwork will integrate, at European level,existing national research capacities and willestablish Europe as the international scien-tific leader in the relevant R&D domains.Co-ordinator: Stelios Orphanoudakis,FORTH.

Foundations, Software Infrastructuresand Applications for Large-ScaleDistributed, Grid and Peer-to-PeerTechnologies — CoreGRIDObjectives: The network will gathertogether European institutions (mainlyacademics but with support fromEuropean industry) carrying out R&Dactivit ies in the area of large scaledistributed computing using Grid andPeer-to-Peer (P2P) technologies.Co-ordinator: Michel Cosnard, INRIA.

Image and Video Understanding:Extracting Semantics from VisualData — XViSObjectives: The network will act as avirtual lab in which the main actors willoutline an ambitious vision for the futureand prepare a corresponding roadmap.Support will be sought on a pan-Europeanscale in order to establish the EU as theinternational scientific leader in thisdomain. Closer collaboration among themany European groups that are active inthis area will be fostered by stimulating theexchange of results and ideas, by maxi-mizing the dissemination of information, bypooling complementary expertise throughjoint research initiatives and by creatingfocussed task forces (either virtually orphysically).Co-ordinator: Eric Pauwel, CWI.

Information Systems for Environ-mental Degradation and DisasterMonitoring and Management – ISEMObjectives: To conduct on a Europeanscale a cooperative research effortdefining knowledge-based systems withgeneric components and applicationmodules dedicated to environmentaldegradation and disaster monitoring andmanagement.Co-ordinator: Isabelle Herlin, INRIA.

Intelligent Environments of Use – In-EUObjectives: To mobilize Europe’s R&Dcapabilities in those fields that constitutethe discipline of Human ComputerInteraction. A range of short-, medium- andlong-term activities aimed at providing anew basis for designing and developinguser interface software and technologieswill be pursued in order to empower citi-zens’ interaction with distributed andcontext-aware user environments in theemerging Information Society.Co-ordinator: Costantine Stephanides,FORTH.

Semantic WebObjectives: The aim of this proposal isto establish Europe as the internationalscientific leader in the next-generationSemantic Web.Co-ordinator: Julius Stuller, CRCIM.

ERCIM News No. 51, October 2002

Articles in this SectionIntroduction

7 A Few Words about the Semantic Web and its Development in the ERCIM Institutesby Jérôme Euzenat, INRIA

Semantic Web Initiatives

9 The Semantic Web lifts offby Tim Berners-Lee and Eric Miller

11 Semantic Web Technologies in Europe’s IST Programme 1998 - 2002by Hans-Georg Stork, European Commission

13 Activities of INTAP Semantic Web Committee in Japanby Tomihiko Kojima, INTAP, Japan

14 OntoWeb: The Thematic Network for the Semantic Webby Ying Ding and Dieter Fensel, Vrije Universiteit Amsterdam, The Netherlands

Tools and Experiments

14 Semantic Web Advanced Development in Europeby Brian Matthews, Michael Wilson, CLRC; and Dan Brickley, W3C

16 Calendars, Schedules and the Semantic Webby Terry R. Payne, University of Southampton and Libby Miller University of Bristol

18 The ICS-FORTH RDFSuite: High-level Scalable Tools for the Semantic Webby Sofia Alexaki, Nikos Athanasis, Vassilis Christophides, Greg Karvounarakis,Aimilia Maganaraki, Dimitris Plexousakis and Karsten Tolle, FORTH

19 Corporate Semantic Websby Rose Dieng-Kuntz, INRIA

21 Joint Research Centre of the EC to use Semantic Web for New Privacy Agentby Giles Hogben and Marc Wilikens, Joint Research Center, Italy

23 Exmo: Assessing Semantic Properties of Knowledge Transformationsby Jérôme Euzenat, INRIA

Metadata, Ontologies and Information Retrieval

24 CORES — A Forum on Shared Metadata Vocabulariesby Tom Baker, Fraunhofer-Gesellschaft

25 Towards Syntax-Independent B2Bby Birgit Hofreiter, Christian Huemer and Werner Winiwarter, University of Vienna

26 Extended Faceted Taxonomies for Web Catalogsby Yannis Tzitzikas, Nicolas Spyratos, Panos Constantopoulos and Anastasia Analyti, FORTH

28 Migrating Thesauri to the Semantic Webby Michael Wilson and Brian Matthews, CLRC

29 Semantic Characterisation of Links and Documentsby Silvia Martelli and Oreste Signore, ISTI-CNR

31 An Ontology-Oriented Solution for Knowledge-Intensive Organizationsby Emanuele Della Valle and Maurizio Brioschi, CEFRIEL, Politecnico di Milano, Italy

32 Combining XML and Description Logics for Describing and Querying Documentsby Rim Alhulou and Amedeo Napoli, LORIA, France

33 Harmonise: An Ontology-Based Approach for Semantic Interoperabilityby Michele Missikoff, IASI-CNR

34 Reducing the Complexity of Information Retrieval by Weighted Queryingby Peter Vojtáš and Jan Vinafi, P. J. Šafárik University, Košice (SRCIM)

Multimedia and Adaptation

36 Hypermedia Presentation Generation on the Semantic Webby Lynda Hardman, CWI

37 Profile-Based Adaptability in the Semantic Webby Santtu Toivonen, VTT

38 Integrating Multimedia Components into a Semantic Webby Benjamin Jung, Fraunhofer-Gesellschaft, and Lyndon JB Nixon, TCD

39 Facilitating Autonomous Active Learning in Web-Based Course Environmentsby Claus Pahl, Dublin City University, Ireland

Others

41 Concepts on Enriching, Understanding and Retrieving the Semantics on the Webby Arne Sølvberg, Sari Hakkarainen, Terje Brasethvik, Xiaomeng Su,Mihhail Matskin and Darijus Strasunskas, NTNU

42 Negotiating on Your Own Terms!by Declan O’Sullivan and Vincent Wade, TCD

SPECIAL THEME: SEMANTIC WEB

by Jérôme Euzenat

A first illustration of the direction of the Semantic Web consistsin noting that the Web as it currently exists is very difficult tosearch. For instance, looking for a book about Agatha Christieis not easy, since many current search engines discard the word‘about’ as meaningless, and may then return a plethora of pagesreferring to books by Agatha Christie. If we want machines toperform effective document searches, we must help them a bitby telling them what documents mean. The correct identifica-tion of relational information (such as relating a Book object toa Person object named ‘Agatha Christie’ using the ‘about’ or‘author’ link) would be a step towards a more accurate search.It is thus natural that the first language made for the SemanticWeb, RDF (Resource Description Framework), emphasisesthese relationships.

The idea of the Semantic Web (a term coined by Tim Berners-Lee, the creator of the Web) involves annotating documentswith ‘semantic markup’, that is, markup that is not interpretedfor display but rather as an expression of document content.This is often described as ‘a Web for machines’ as opposed to aWeb to be read by humans. This idea was experimented withon a small scale during the nineties, both by the SHOE system,developed at the University of Maryland, and Ontobroker,developed at the University of Karlsruhe.

The development of the Semantic Web took off with theDARPA Agent Markup Language (DAML) initiative in theUSA. Its goal was to provide the semantic markup languagesuccessor to SHOE. Shortly after, European researcherscreated the Ontoweb thematic network in order to federate theresearch in these fields, and W3C launched its long-awaitedSemantic Web activity that took over the development of RDFand its extensions. These had previously been advanced by thetwo other groups. Research activities for building the SemanticWeb are now active all over the world, and are central to the‘knowledge technologies’ area of the European Union’s 6thframework program. We invited several major players inSemantic Web research (W3C, the European Commission,INTAP and the Ontoweb network) to present their current andfuture activities in the domain.

The applications of the Semantic Web are limited only by ourimagination. A more elaborate scenario for the Semantic Webinvolves not only the facility for improved searching, but alsofor selecting, assembling and triggering services found on the

Web, matching interest profiles and resource descriptions,automatically rescheduling and reassembling services upon anunexpected event, notifying customers of a schedule change ontheir favourite communication medium or mining biomoleculardatabanks at night.

For instance, the ‘travel agent scenario’ deals with a softwareagent able to plan a complex trip involving several forms oftransportation, hotel stays, and conference and entertainmentregistration by resolving heterogeneous constraints. Thesecould include using trusted hotels in preferred areas, takingeating requirements into account, finding connections, usingonly prescribed airlines and frequent-flyer-affiliated car rental,and minimising the total cost of the trip. Such a scenario neces-sitates four main capabilities on the part of the agent:1.a network of well-annotated resources, eg plane schedules

and fares, hotel locations and facility descriptions, etc2.a reservoir of available knowledge (or an ontology), eg a bus

is a means of transportation, Sardinia is in Italy, etc3.a description of user preferences, eg food preferences,

accommodation needs, frequent flyer programs, agendas, etc4. inference capabilities, eg taxonomical inference, temporal

inference, trust propagation, etc.

In short, this scenario involves doing what a very good assistantcould do, but in a more systematic and provable way (thoughperhaps in a less flexible and agreeable manner). However, theSemantic Web should not be confused with what has some-times been called ‘strong artificial intelligence’ for threereasons. Firstly, there is no anthropomorphic claim to theSemantic Web – instead it is intended to complement humansin areas where they do not perform very well (ie, dealingquickly with large amounts of information, working continu-ously, analysing large texts for certain pieces of information,etc). Secondly, it takes into account many lessons learnt fromthe Web; that is, being large, inconsistent and spread, SemanticWeb applications must be scalable, robust and decentralised.Lastly, it must be adapted to the context in which it will evolve.

The research aspects of the Semantic Web were debated at arecent ERCIM strategic workshop that was held in Sophia-Antipolis last year. It identified several important researchdirections to be investigated: • identification and localisation, for the correct and universal

addressing of resources

A Few Words about the Semantic Web...





• relationships between semantic models, for adequatelytransforming and interpreting the content of the SemanticWeb

• tolerant and safe reasoning, for overcoming the inherentheterogeneity and inconsistency of an open world, and

• facilitating Semantic Web adoption, by imagining originalknowledge-capture means and adequate growth models.

From this, it is evident that a number of fields can contribute tothe development of the Semantic Web. Many groups are nowapplying their knowledge at various levels, from how to putcategory information on Web pages to how computers can berelated to meaning. As a matter of fact, one of the goodfeatures of the Semantic Web is that it has encouraged theclose cooperation of many people from different backgrounds.We can already see structured markup developers talkinglogic with knowledge representation researchers, databaseengineers talking protocols with multimedia designers, elec-tronic commerce managers starting to talk to agent developers.Of course, the Semantic Web primarily takes advantage of thetechniques developed on the Web and in knowledge represen-tation, but other fields also have important contributions tomake. These include document management for manipulatingcomplex markups, databases for dealing efficiently with asser-tions, digital libraries for metadata and search schemes, logicsfor crafting language semantics and reasoning systems, inter-face design for offering new views of these marked docu-ments, and agents for gathering knowledge and negotiatingover the network.

This variety of topics in research and experimentation is wellillustrated by the work presented in this special issue. TheERCIM people are working hard on the development of theSemantic Web, and we have organised their presentations intothree sections: • ‘Tools and Experiments’ gathers projects that are devel-

oping tools for manipulating and experimenting withsemantic markup

• ‘Metadata, Ontologies and Information Retrieval’ is dedi-cated to projects on the design and alignment of vocabu-laries for use in marking up and retrieving documents andproducts

• ‘Multimedia and Adaptation’ deals with the modular designof multimedia documents, and the appeal they hold in termsof individual adaptation and retrieval as well as profiledassembly.

We are now at a moment in which the players are mobilised inorder to provide the best languages, architectures and tools forexperimenting with the concepts of the Semantic Web. In thisspecial issue, you have the opportunity to read all about it.

...and its Development in the ERCIM Institutes

Links:Semanticweb.org: http://www.semanticweb.orgW3C/SW activity: http://www.w3.org/2001/sw/Ontoweb: http://www.ontoweb.orgDAML: http://www.daml.orgThe International Semantic Web Conferences:http://iswc.semanticweb.org/

Further reading:Tim Berners-Lee, James Hendler, Ora Lassila, The Semantic Seb, Scientific American 284(5):35-43, 2001, http://www.scientificamerican.com/2001/0501issue/0501berners-lee.html

Isabel Cruz, Stefan Decker, Jérôme Euzenat, Deborah McGuinness (eds.), The Emerging Semantic Web: Selected Papers from the First Semantic Web Working Symposium, IOS press, Amsterdam, 2002, 300pp, http://www.inrialpes.fr/exmo/papers/emerging/

Ian Horrocks, James Hendler (eds.), The Semantic Web - ISWC 2002, Lecture Notes in Computer Science 2342, Springer-Verlag, 2002 http://link.springer.de/link/service/series/0558/tocs/t2342.htm

Dieter Fensel, James Hendler, Henry Lieberman, Wolfgang Wahlster (eds.), Spinning the Semantic Web, The MIT press, Cambridge, 2002, to appear

Jérôme Euzenat (ed.), Research Challenges and Perspectives of the Semantic Web,http://www.ercim.org/EUNSF/semweb.html

Please contact:Jérôme Euzenat, INRIATel: +33 476 61 53 66E-mail: [email protected]



The Semantic Web is an extension of thecurrent Web in which information isgiven well-defined meaning, enablingcomputers and people to work in bettercooperation. The W3C Semantic WebActivity, in collaboration with a largenumber of researchers and industrialpartners, is tasked with defining stan-dards and technologies that allow data onthe Web to be defined and linked in away that it can be used for more effectivediscovery, automation, integration, andreuse across applications. The Web willreach its full potential when it becomesan environment where data can be sharedand processed by automated tools as wellas by people.

How might this be useful? Suppose youwant to compare the price and choice offlower bulbs that grow best in your zipcode, or you want to search online cata-logs from different manufactures forequivalent replacement parts for a Volvo740. The raw information that mayanswer these questions, may indeed beon the Web, but it’s not in a machine-usable form. You still need a person todiscern the meaning of the informationand its relevance to your needs.

The Semantic Web addresses thisproblem in two ways. First, it will enablecommunities to expose their data so thata program doesn’t have to strip theformatting, pictures and ads from a Webpage to guess at the relevant bits of infor-mation. Secondly, it will allow people towrite (or generate) files which explain —to a machine — the relationship betweendifferent sets of data. For example, onewill be able to make a ‘semantic link’between a database with a ‘zip-code’column and a form with a ‘zip’ field thatthey actually mean the same thing. This

will allow machines to follow links andfacilitate the integration of data frommany different sources.

This notion of being able to semanticallylink various resources (documents,images, people, concepts, etc) is animportant one. With this we can begin tomove from the current Web of simplehyperlinks to a more expressive semanti-cally rich Web, a Web where we canincrementally add meaning and expressa whole new set of relationships(hasLocation, worksFor, isAuthorOf,hasSubjectOf, dependsOn, etc) amongresources, making explicit the particularcontextual relationships that are implicitin the current Web. This will open newdoors for effective information integra-tion, management and automatedservices.

How is it being developed?There are two places to look forSemantic Web progress: from theground up, in the infrastructural andarchitectural work coordinated by theW3C, and from top down, in application-specific work by those leveragingSemantic Web technologies in variousdemonstrations, applications and prod-ucts. This article provides an introduc-tion to both views with a specific focuson those areas in which the W3C isdirectly involved.

Enabling StandardsUniform Resource Identifiers (URIs) area fundamental component of the currentWeb and are a foundation of theSemantic Web. The Extensible MarkupLanguage (XML) is also a fundamentalcomponent for supporting the SemanticWeb. XML provides an interoperablesyntactical foundation upon which the

more important issue of representingrelationships and meaning can be built.URIs provide the ability for uniquelyidentifying resources as well as relation-ships among resources. The ResourceDescription Framework (RDF) family ofstandards leverages URIs and XML toprovide an stepwise set of functionalityto represent these relationships andmeaning.

The W3C Semantic Web Activity’scharter is to serve a leadership role in thedesign of specifications and the open,collaborative development of technolo-gies focused on representing relation-ships and meaning. The base level of theRDF family of standards is a W3CRecommendation. The RDF CoreWorking Group is in the process offormalizing the original RDF Model andSyntax Recommendation whichprovides a simple yet powerful frame-work for representing information in theWeb. Building on this work, the group isadditionally defining a simple means fordeclaring RDF Vocabularies. RDFVocabularies are descriptive terms (egService, Book, Image, title, description,rights, etc) that are useful to communi-ties recoding information in a way thatenables effective reuse, integration andaggregation of data. Additional deliver-ables include a precise semantic theoryof these standards that will supportfuture work, as well as a primer designedto provide the reader with a basic under-standing of RDF and its application.

Simple data integration, aggregation andinteroperability are enabled by thesebase level RDF standards. An increasingneed for interoperability at a moreexpressive descriptive level is alsodesired. The Web Ontology Working

Many researchers at ERCIM Institutes are aware that this is an exciting time tobe involved in work done at the World Wide Web Consortium (W3C). ScalableVector Graphics, Web Services, and the Semantic Web are but a few of the W3CActivities attracting media attention. This article focuses on the W3C’s SemanticWeb Activity and recent developments in the Semantic Web community. Althoughit is difficult to predict the impact of such a far-reaching technology, currentimplementation and signs of adoption are encouraging and developments infuture research areas are extremely promising.

The Semantic Web lifts offby Tim Berners-Lee and Eric Miller



Group is charted to build upon the RDFCore work a language for defining struc-tured, Web-based ontologies. Ontologiescan be used by automated tools to poweradvanced services such as more accurateWeb search, intelligent software agentsand knowledge management. Webportals, corporate website management,intelligent agents and ubiquitouscomputing are just some of the identifiedscenarios that helped shaped the require-ments for this work.

Advanced DevelopmentJust as the early development of the Webdepended on code modules such aslibwww, W3C is devoting resources tothe creation and distribution of similarcore components that will form the basisfor the Semantic Web. The W3CSemantic Web Advanced Development(SWAD) initiatives are designed to workin collaboration with a large number ofresearchers and industrial partners andstimulate complementary areas of devel-opment that will help facilitate thedeployment and and future standardswork associated with the Semantic Web.

SWAD DAMLThe purpose of the SWAD DAMLproject is to contribute to the develop-ment of a vibrant, ubiquitous SemanticWeb by building critical Semantic Webinfrastructure and demonstrating howthat infrastructure can be used byworking, user-oriented applications.

SWAD DAML is designed to build onthe DARPA Agent Markup Language(DAML) infrastructure to provide aninterchange between two or moredifferent applications. The first involvesstructured information manipulationrequired to maintain the ongoing activi-ties of an organization such as the W3C.These include access control, collabora-tive development, and meeting manage-ment. The second application is focusedon the informal and often heuristicprocesses involved in documentmanagement in a personalized informa-tion environment. Integrated into bothenvironments will be tools to enableauthors to control terms under whichpersonal or sensitive information is usedby others, a critical feature to encouragesharing of semantic content.

SWAD-EuropeSWAD-Europe will highlight practicalexamples of where real value can beadded to the Web through Semantic Webtechnologies. The focus on this initiativeis on providing practical demonstrationsof how the Semantic Web can addressproblems in areas such as: site maps,news channel syndication, thesauri, clas-sification, topic maps, calendaring,scheduling, collaboration, annotations,quality ratings, shared bookmarks,Dublin Core for simple resourcediscovery, Web service description anddiscovery, trust and rights managementand how to effectively and efficientlyintegrate these technologies together.

SWAD-Europe will additionallyconcentrate on exploratory implementa-tion and pre-consensus design in areassuch as querying, and the integration ofmultiple Semantic Web technologies. Itshall provide valuable input and experi-ences to future standards work.

SWAD SimileW3C is additionally working with HP,MIT Libraries, and MIT’s Lab forComputer Science on Simile, whichseeks to enhance interoperability amongdigital assets, schemas, metadata, andservices across distributed individual,community, and institutional stores andacross value chains that provide usefulend-user services by drawing upon theassets, schemas, and metadata held insuch stores. Simile will leverage andextend DSpace, enhancing its support forarbitrary schemas and metadata,primarily though the application of RDFand Semantic Web techniques. Theproject also aims to implement a digitalasset dissemination architecture basedupon Web standards, enabling servicesto operate upon relevant assets, schemas,and metadata within distributed stores.

The Simile effort will be grounded byfocusing on well-defined, real-world usecases in the libraries’ domain. Sinceparallel work is underway to deployDSpace at a number of leading researchlibraries, we hope that such an approachwill lead to a powerful deploymentchannel through which the utility andreadiness of Semantic Web tools andtechniques can be compellingly demon-strated in a visible and global community.

SWAD OxygenThe MIT/LCS Oxygen project isdesigned to enable pervasive, human-centered computing through a combina-tion of specific user and system tech-nologies. Oxygen’s user technologiesdirectly address human needs. Speechand vision technologies enable us tocommunicate with Oxygen as if we’reinteracting with another person, savingmuch time and effort. Automation, indi-vidualized knowledge access, andcollaboration technologies help usperform a wide variety of tasks in theway we like to do them. In Oxygen,these technologies enable the formationof spontaneous collaborative regions thatprovide support for recording, archiving,and linking fragments of meetingrecords to issues, summaries, keywords,and annotations.

The Semantic Web is designed to fostersimilar collaborative environment andthe W3C is working with projectOxygen to help support this goal. Theability for “anyone to say anything aboutanything” is an important characteristicof the current Web and is a fundamentalprinciple of the Semantic Web. Knowingwho is making these assertions isincreasingly important in trusting thesedescriptions and enabling a ‘Web ofTrust’. The Annotea advanced develop-ment project provides the basis forasserting descriptive information,comments, notes, reviews, explanations,or other types of external remarks to anyresource. Together with XML digitalsignatures, the Annotea project willprovide a test-bed for ‘Web-of-Trust’Semantic Web applications.

Applications - spinning upwardThough not the focus of this article, thedeployment of RDF-based technologiesis increasingly significant. The W3CSemantic Web Activity hosts the RDFInterest Group, which coordinates publicimplementation and shares deploymentexperiences of these technologies.Arising out of RDF Interest Groupdiscussions are several public issue-specific mailing lists, including RDF-based calendar and group schedulingsystems, logic-based languages, queriesand rules for RDF data and distributedannotation and collaboration systems.These discussion groups are designed to



focus on complementary areas of interestassociated with the Semantic WebActivity, each of which fosterscooperation and collaboration amongindividuals and organizations workingon related Semantic Web technologies.

In addition to these Interest Group liststhere are a variety of domain specificcommunities who are using RDF/XMLto publish their data on the Web. Thesenotably include the Dublin CoreMetadata Initiative, the IMS GlobalLearning Consortium vocabularies forfacilitating online distributed learning,XMLnews, PRISM, the RDF SiteSummary (RSS 1.0) for supporting newssyndication, Musicbrainz for catalogingand cross-referencing music, andCreative Commons for supporting adigital rights description to name but afew. The Topic Map (XTM) communityhas been finding increasing synergy withthe RDF data model.

Early commercial adopters such asAdobe’s eXtensible Metadata Platform(XMP), for example, leverageRDF/XML to enable more effectivemanagement of digital resources. Adobeapplications and workflow partnersthrough XMP can leverage the power ofRDF/XML to provide a standardizedmeans for supporting the creation,

processing, and interchange of documentmetadata across publishing workflows.This in-turn reduces cost and makes formore effective management of digitalresources possible both within andacross organizational boundaries.

New Things opening upThe most exciting thing about theSemantic Web is not what we canimagine doing with it, but what we can’tyet imagine it will do. Just as globalindexes, and Google’s algorithms werenot dreamed of in the early Web days,we cannot imagine now all the newresearch challenges and exciting productareas which will appear once there is aWeb of data to explore. Many existingfields for knowledge representation anddata management have typically madeassumptions regarding a conceptually orphysically centralized system, and assuch their application to the SemanticWeb is not straightforward. Given amass of rules relating data in differentvocabularies, and an unbounded set ofdatasets in different vocabularies, whatalgorithms will efficiently resolvegeneral queries? What conventions forthe storage of tips and pointers will allowdata to be reused and converted automat-ically? What techniques will allow asystem to operate securely whileprocessing very diverse data from

untrusted agents? How can one represent— and then implement — personalprivacy in such a world?

The Semantic Web starts as a simplecircles-and-arrows diagram relatingthings, which slowly expands andcoalesces to become global and vast. TheWeb of human-readable documentsspawned a social revolution. TheSemantic Web may in turn spawn a revo-lution in computing. In neither case did achange occur in the power of one personor one computer, but rather a dramaticchange in the role they can play in theworld, by being able to find out almostanything virtually immediately.

For more information on the SemanticWeb, including additional projects, prod-ucts, efforts and future directions checkout the Semantic Web home page.

Link:http://www.w3c.org/2001/sw/

Please contact:Tim Berners-Lee, Director of W3CE-mail: [email protected]

Eric MillerW3C Semantic Web Activity LeadE-mail: [email protected]

The European Commission’s Infor-mation Society Technologies (IST)Programme was designed in 1997-1998as part of the 5th Framework Programmefor R&D in Europe, covering the period1998 - 2002. One of its Key Action III(‘Multimedia Content and Tools’)modules was entitled ‘Informationaccess, filtering, analysis and handling’(IAF for short). Its objectives were to

support the development of “...advancedtechnologies for the management ofinformation content to empower the userto select, receive and manipulate ... onlythe information required when facedwith an ever increasing range of hetero-geneous sources.” These technologiesshould lead to “... improvements in thekey functionalities of large-scale multi-media asset management systems

(including the evolution of the WorldWide Web) to support the cost-effectivedelivery of information services and theirusage.” (Quoted from http://www.cordis.lu/ist/b-oj-en5.htm#KA3).

Only a few months before the first ISTCall was launched in early 1999, TimBerners-Lee had published his informal‘Semantic Web Road Map’ note on the

by Hans-Georg Stork

The European Commission supports a broad range of research and developmentactivities related to the Semantic Web. This support is expected to continue inthe forthcoming 6th Framework Programme.

Semantic Web Technologiesin Europe’s IST Programme 1998 - 2002



W3C website, outlining possible futuredirections for the evolution of the WorldWide Web. These ideas, partly based onprevious content and resource descrip-tion activities, had met with growingenthusiasm of researchers and devel-opers worldwide, both in academia andin industry. They provided a new visionfor integrating efforts that had beenongoing for some time in many R&Dcommunities, involving specialists invarious computer science disciplines.These efforts were aimed at capturingthe semantics of digital content of allsorts, and at devising ways of actingsensibly upon the formal knowledgerepresentations thus gained.

Picking up on this vision the EuropeanCommission dedicated a specific actionline of its IST Work Programme 2001 to‘Semantic Web Technologies’ (ActionLine III.4.1), thereby underlining theimportance (in terms of research chal-lenges and expected impact) of ‘seman-tics issues’ for achieving the declaredgoals of the IAF module of IST.

It offered four broad interrelated R&Dtracks as an orientation for submittingproject proposals:• creating a usable formal framework in

terms of formal methods, models,languages and corresponding tools forsemantically sound machine-process-able resource description

• fleshing out the formal skeletons bydeveloping and applying techniquesfor knowledge discovery (in databasesand text repositories), ontologylearning, multimedia content analysis,content-based indexing, etc

• acting in a semantically rich environ-ment, performing resource and servicediscovery, complex transactions,semantic search and retrieval, filteringand profiling, supporting collaborativefiltering and knowledge sharing, etc

• making it understandable to peoplethrough device-dependent informationvisualisation, semantics-based andcontext-sensitive navigation andbrowsing, semantics-based dialoguemanagement, etc.

This agenda provided some continuitywith respect to previous Key Action IIIactivities (notably on ‘media representa-tion and access’ and digital libraries) and

activities supported by other IST depart-ments (for instance under action line‘Methods and tools for intelligence andknowledge sharing’ of Key Action IV -Essential Technologies and Infrastruc-tures, or under the Open Domain of FET— Future and Emerging Technologies).

But it also provided a sharper focus onthe problems of creating and usingknowledge representations, in the

context of large-scale distributedsystems, such as the World Wide Web.

This focus was largely retained in WorkProgramme 2002 as part of Key ActionIII’s ‘Preparing for future research activ-ities’ action line (AL III.5.2). Moreover,Work Programme 2002, in one of its‘Cross Programme Activities (CPA)’,took account of a new trend that hassurfaced over the last couple of years:

COGCorporate Ontology Grid http://www.cogproject.org/

ESPERONTOApplication Service Provision of SemanticAnnotation, Aggregation, Indexing andRouting of Textual, Multimedia, andMultilingual Web Contenthttp://www.isoco.com/isococom/whitepapers/files/ SemanticWeb-whitepaper-137.pdf

FF-POIROTFinancial Fraud Prevention-OrientedInformation Resources using OntologyTechnologyhttp://www.starlab.vub.ac.be/research/projects/default.htm#Poirot

GRACEGrid Search and Categorisation EngineIST-2001-38100

IBROWAn Intelligent Brokering Service forKnowledge-Component Reuse on theWorld Wide Webhttp://www.swi.psy.uva.nl/projects/ibrow/

InDiCoIntegrated Digital ConferencingIST-2001- 34306

MONETMathematics on the Nethttp://monet.nag.co.uk/

MOSESA modular and Scalable Environment for the Semantic WebIST-2001-37244

On-To-KnowledgeContent-Driven Knowledge Managementthrough Evolving Ontologieshttp://www.ontoknowledge.com

ONTO-LOGGINGCorporate Ontology Modelling andManagement Systemhttp://www.ontologging.com

SCULPTEURSemantic and Content-Based MultimediaExploitation for European BenefitIST-2001-35372

SEWASIESemantic Webs and Agents in Integrated Economieshttp://www.sewasie.org/

SPACEMANTIXCombining Spatial and SemanticInformation in Product DataIST-2001-34159

SPIRITSpatially-Aware Information Retrieval on the Internethttp://www.cs.cf.ac.uk/department/posts/SPIRITSummary.pdf

SWAD-EuropeW3C Semantic Web AdvancedDevelopment for Europehttp://www.w3.org/2001/sw/Europe/

SWAPSemantic Web and Peer-to-Peerhttp://swap.semanticweb.org/

SWWSSemantic-Web-Enabled Web Serviceshttp://www.cs.vu.nl/~swws/

VICODIVisual Contextualisation of Digital ContentIST-2001-37534

WIDESemantic-Web-Based InformationManagement and Knowledge-Sharing forInnovative Product Design and Engineeringhttp://www.cefriel.it/topics/research/default.xml?id=75&_euro=1&old=1&dc=1&tid=27

WISPERWorldwide Intelligent Semantic PatentExtraction & RetrievalIST-2001-34407

WonderWebOntology Infrastructure for the Semantic Webhttp://wonderweb.semanticweb.org

Some IST Projects on Semantics in Distributed Systems(‘Semantic Web Technologies’)



INTAP is an association authorised bythe Ministry of Economy, Trade andIndustry (or METI) of Japan, establishedin 1985. The Association conductsR&D, surveys, and publicity activitiesfor promoting interoperability of infor-mation and communication technolo-gies. The Association stimulates theprogress of information processing andrelated technologies, which shouldcontribute to the creation of a sound andhealthy information society.

INTAP Semantic Web CommitteeINTAP has established the SemanticWeb Committee, based on recognitionthat the Semantic Web is a critical tech-nology that will affect widely economic

activities and social life. INTAP drivesforward research and developmentrelated to the Semantic Web and itspromotion in Japan through theCommittee, in cooperation with KeioUniversity and the World Wide WebConsortium (W3C).

A brief history of the Semantic WebCommittee:• established in 2001• Semantic Web Conference (October

2001, at Keio University)• Semantic Web Workshop (December

2001, at DBWeb2001 in Kyoto)• presentation in panel session by the

committee chairman, Noboru Shimizuat ISWC2002 (June 2002)

• Semantic Web special report of thetechnical magazine issued by theInformation Processing Society ofJapan (IPSJ) in July 2002:What is the Semantic Web?; Meta-Data on the Semantic Web and itsUsage; Web Ontology language forthe Semantic Web; Tools for theSemantic Web; Application Systemson the Semantic Web

• Semantic Web Conference 2002 in18th September 2002.

Members of the CommitteeThe committee is chaired by Mr. NoboruShimizu of NEC Corporation and techni-cally led by Dr. Nobuo Saito, VicePresident of Keio University and Dr.

by Tomihiko Kojima

The Interoperability Technology Association for Information Processing (INTAP)drives forward researches and developments related to the Semantic Web andits promotion in Japan.

Activities of INTAP Semantic Web Committee in Japan

the application of Grid technologies to“knowledge discovery in ... largedistributed datasets, using cognitivetechniques, data mining, machinelearning, ontology engineering, informa-tion visualisation, intelligent agents...(quoted from WP2002, CPA9)”, allmore or less directly pertinent to theSemantic Web vision.

Calls for submission of proposals tothese action lines were published in July(AL III.4.1) and November (AL III.5.2& CPA9) 2001, respectively (Calls 7 and8). Both calls met with considerableinterest in relevant R&D communitiesacross Europe and drew altogethernearly one hundred submissionsinvolving several hundred participatingorganisations. They resulted in a signifi-cant growth (by 17 projects) of a port-folio of projects that are all poised tocontribute in one way or another, tomaking the Semantic Web a reality (seethe references in the box, including a fewearlier and concurrent projects in KeyActions II, IV and in FET).

While at the time of writing the newprojects have only just commenced, someof the older ones have already producednoteworthy results. It may suffice tomention project On-To-Knowledge(http://www.ontoknowledge.com), whichhas been one of the birthing grounds ofOWL, the proposed new Web ontologylanguage, currently under discussion atthe W3C.

Moreover, in recognition of the centralrole ontologies are likely to play inbuilding the ‘Semantic Web’, theEuropean Commission, through its ISTProgramme, supports the ‘ThematicNetwork’ OntoWeb, a platform forfostering collaboration between industryand academia, on creating a ‘semanticinfrastructure’ for applications in manydifferent areas (e-business, Webservices, multimedia asset management,community webs, etc). ThroughOntoWeb, European researchers andpractitioners also have an opportunity tomake more targeted contributions to

international standardisation activitiesand to the W3C process.

It is expected that EU support ofSemantic-Web-related R&D willcontinue under the forthcoming 6thFramework Programme within thebroader context of ‘KnowledgeTechnologies’, as part of the ‘PriorityThematic Area’ IST - InformationSociety Technologies.

The views expressed in this article arethose of the author and do not neces-sarily engage the European Commission.

Links: http://www.cordis.lu/fp6/http://www.cordis.lu/ist/fp6/workshops.htm http://www.cordis.lu/ist/ka3/iaf/iaf_workshop.htm

Please contact: Hans-Georg StorkEuropean Commission, DG Information SocietyE-mail: [email protected]



Tatsuya Hagino, Professor of the sameUniversity. Dr. Saito is working forW3C as the Director and DeputyDirector of W3C. The committeemembers are comprised of NECCorporation, Hitachi Ltd, MitsubishiElectric Corporation, Oki ElectricIndustry Co Ltd, Sumitomo ElectricIndustries, Ltd, Toshiba Corporation,IBM Japan Ltd, Matsushita ElectricIndustrial Co Ltd, Fujitsu Ltd, FujitsuLaboratories Ltd, NEC Software KyushuLtd, NEC Planning Research Ltd, andNippon Telegraph and TelephoneCorporation.

Activities of the Committee• Survey on trends on Semantic Web

technology in Europe and the US:RDF, OWL, DAML, DAML-S, RSS,Dublin Core, etc.

• Research on the successful story ofSemantic Web applications, such asnext-generation e-government.

Blueprints• Development of Japanese Content

Description System using RDF:We will develop a Japanese ContentDescription System using RDF thatcan automatically generate RDF meta-data on a certain Japanese contentbased on pattern formats which areused to identify particular letter stringsas metadata for that content. We willalso develop a Japanese RDF MetadataStandard to describe summaries ofvarious Japanese contents.

• Development of IntegratedManagement System of DistributedInformation (License Repository):Based on the assumption that theInternet is a huge repository, we willdevelop an Integrated Management

System of Distributed Information thatenables attached documents of a writtenapplication to be held on an applicant-side system and enables the staff ofadministrative organs who receive theapplication to browse the attacheddocuments when they want to see them.

• Research and Development of IntelligentSearch System using RDF and Ontology:We will research RDF, ontologydescription language including OWLand intelligent agent technology andwe will develop an Intelligent SearchSystem using RDF and Ontologybased on the research.

Link:http://www.net.intap.or.jp/INTAP/s-web/

Please contact:Tomihiko Kojima, INTAP, JapanTel: +81 3 5977 1301 E-mail: [email protected], [email protected]

The goal of the OntoWeb is to bringresearchers and industry together toenable ontologies to improve informa-tion exchange and provide betterservices. Meanwhile, OntoWeb alsoaims to strengthen the European influ-ence on standardisation efforts in areassuch as Web languages (XML, RDF,OWL), upper-layer ontologies, andcontent standards.

OntoWeb: The Hub for a Semantic WebThe World Wide Web is currentlymoving from the first generation Web tothe second generation Web — theSemantic Web. Tim Berners-Lee coinedthe vision of a Semantic Web thatprovides more automated services basedon machine-processable semantics ofdata and heuristics that make use of thesemetadata.

Ontologies, which provide shared andcommon domain theories, are a key assetfor such a Semantic Web. They can beseen as metadata that explicitly representsemantics of data in a machine process-able way. Ontology-based reasoningservices can operationalise this semanticsfor providing various services. Ontologiestherefore play a crucial role in enablingcontent-based access, interoperability, andcommunication across the Web, providingit with a qualitatively new level of service— the Semantic Web. It weaves together anet linking incredibly large parts of thehuman knowledge and complements itwith machine processability.

Currently, there exist five SpecialInterest Groups (SIGs) within OntoWeb:

SIG1 on content standards: SIG1 aimsto coordinate cooperation and participa-

tion in current initiatives related toontology-based content standardizationand content harmonization acrossdifferent standards. For more details,please visit: http://www.ladseb.pd.cnr.it/infor/ontology/OntoWeb/SIGContentStandards.html

SIG2 on ontology language standards:SIG2 focuses on ontology language stan-dards through coordinating cooperation,communication, and participation inrelated initiatives and relevant standard-ization efforts. For more details, pleasevisit: http://www.cs.man.ac.uk/~horrocks/OntoWeb/SIG/

SIG3 on enterprise-standard ontologyenvironments: The mission of this SIG isto support a dialog to develop qualityrequirements for enterprise-standardontology environments and steer future

by Ying Ding and Dieter Fensel

OntoWeb: Ontology-based information exchange for knowledge managementand electronic commerce is an EU-funded three-year IST Thematic Network projectwithin the 5th Framework Programme.

OntoWeb: The Thematic Network for the Semantic Web


SPECIAL THEME: SEMATIC WEB

ontology development towardsaddressing these requirements. For moredetails, please visit: http://delicias.dia.fi.upm.es/ontoweb/sig-tools/index.html

SIG4 on industrial applications: SIG4offers an international forum for theexchange of ideas, practical experiences,and advances in research into industryand business applications of ontologyand Semantic Web technology, especiallythe Web services powered by SemanticWeb technology. For more details, pleasevisit its homepage: http://www.cs.vu.nl/~maksym/sig4/sig4.html

SIG5 on language technology inontology development and use: TheMotivations of SIG5 is to provide theconnections between language tech-nology and ontologies. Being the plat-form for cooperation and standards,SIG5 conveys the multidisciplinaryexpertise between natural language tech-nology and Semantic Web. For moredetails, please visit: http://dfki.de/~paulb/ontoweb-lt.html

OntoWeb is the European focal point forbringing together activities in the area ofontology-based methods and tools for

the Semantic Web. It will significantlysecure a competitive position in theemerging business field by:• envisioning and describing the current

state-of-the-art in ontology-relatedfields in Europe and world-wide,providing a strategic guide to industrialand commercial applications

• co-operating with different standard-isation bodies to promote the develop-ment of ontology-based contentstandardisation and ontology languagestandardisation

• providing non-discriminatory accessto services by individuals and busi-nesses and aiding identification andextraction of information in a rapidlygrowing network.

• organising dissemination workshops,special interest groups, a scientificjournal, and training or educationcourses with special emphasis onSemantic-Web-Based applications, e-commerce, knowledge managementand information integration. Itaddresses aspects of knowledgetransfer between academia andindustry for the emerging ontology-based technologies, enabling competi-tive advantages in these fast devel-oping markets.

OntoWeb currently has about 120members from Europe, America, andAsia. The ratio of academic and commer-cial members is roughly 2:1. We invitecompanies, research institutes, and univer-sities active in the Semantic Web area tojoin the OntoWeb project. We offer theopportunity for project members to:• share the enriched resources of the latest

ongoing research, cutting-edge know-ledge, and innovative technologiesthrough the OntoWeb portal

• attend various OntoWeb conferences,workshops, and seminars to meetpossible future collaborative partners orpotential customers and to exploit newacademic or business opportunities

• join the OntoWeb Special InterestGroups (SIGs).

To become a member, please visithttp://www.ontoweb.org/member.htmLet us work jointly to bring the SemanticWeb to its full potential.

Link:http://www.ontoweb.org/

Please contact: Ying Ding, Dieter FenselVrije Universiteit Amsterdam, The NetherlandsE-mail: {ying, dieter}@cs.vu.nl

by Brian Matthews, Michael Wilson and Dan Brickley

The Semantic Web Activity of the World Wide Web Consortium (W3C) aims toaugment the existing Web architecture by adding communicable informationabout resources which can then be processed by automated agents. This initiativehas attracted much interest in the last few years, and led to the creation of aEuropean project — SWAD-Europe or Semantic Web Advanced Development inEurope which aims to highlight practical examples of where real value can beadded to the Web through the Semantic Web.

Semantic Web Advanced Development in Europe

The initial inspiration for the SWAD-Europe project was a response to ques-tions which frequently arise whenconsidering Semantic Web technology,such as:• How do I use RDF with XML

Schemas?• Web Services with Web Ontologies?• MathML with RDF-rules?

These technology-oriented questionsoften mask an overall application-oriented need. It is not clear which, ifany, of these technologies are mostappropriate to use, nor what the relation-ship between them is. A common themerecurs: technology integration.(Semantic) Web technology has adaunting array of tools, specificationsand techniques — there is too much tochoose from!

SWAD-Europe aims to address theseconcerns by tackling simple things first,to provide real examples of how theSemantic Web can address problems inareas such as: sitemaps, news channelsyndication ; thesauri, classification,topic maps ; calendaring, scheduling,collaboration ; annotations, qualityratings, shared bookmarks ; Dublin Corefor simple resource discovery; Webservice description and discovery; trust



and rights management ; and how to mixall these together. Our general approachis not to focus on the AI-related tech-nology from the higher levels of theSemantic Web architecture, but rather toexploit what has been developed so far tohave it used in practice, therebyanswering some of the queries we hear.SWAD-Europe will concentrate onexploratory implementation and designin pre-consensus areas such as querying,and the integration of multiple SemanticWeb technologies. In this way, we intendto show the utility of the Semantic Webto the European developer community.

An outreach programme is a key compo-nent of the project which will beapproached via showcasing SemanticWeb for industry applications, guidanceon how to integrate Semantic Web,XML and other Web standards, andfacilitating technology maturation, egSemantic-Web-enhanced Web Services.Further, the Semantic Web will beencouraged and deepened for all parts ofthe industrial and academic communitieswithin Europe.

The project will develop scenarios fordemonstrating the practical utility ofSemantic Web technology. We would

like to give developers both a ‘bigpicture’ understanding of the SemanticWeb and a detailed knowledge of thetools and techniques and thus providefeedback for standardisation work. Theproject will be looking at demonstratorprojects and developments in a variety oftechnical areas:• XML / Semantic Web Integration:

Demonstrating that SW technologiesare compatible with the broader familyof XML languages (especially SVGand MathML).

• Web Services Integration: Convergencebetween RDF and SOAP serialisa-tions.

• Scalability, APIs and QueryLanguages: Advances in technicalknowledge with the rationalisation anddocumentation of deployed best prac-tice for APIs and simple querylanguages (eg SquishQL), includingtesting on large triple bases.

• Trust, Annotations and InformationQuality: Specific applications toaugment the Web Services with toolsfrom the Semantic Web.

• Thesauri and Knowledge Management:Simple classification tools to aidknowledge management, with anemphasis on multi-linguality in aEuropean context.

We intend to make technical innovationsin all these areas, and combine them toprovide a practical demonstration of theutility of the Semantic Web.

SWAD-Europe is a collaborationbetween the World Wide WebConsortium in Europe, hosted byINRIA, research institutions at theCentral Laboratory of the ResearchCouncils (CLRC) and the University ofBristol (ILRT), and the companies StiloInternational Plc and Hewlett-PackardLtd.

SWAD-Europe is an EU-funded project,part of the Information SocietyTechnologies (IST-7) programme. Thefirst deliverables are already publicallyavailable at http://www.w3.org/2001/sw/Europe/reports/intro.html).

Links:http://www.w3.org/2001/sw/Europe/http://www.w3.org/2001/sw/

Please contact: Michael Wilson, CLRCE-mail: [email protected]

Dan Brickley, W3CE-mail:[email protected]

Scheduling meetings can be tedious andtime consuming, and involves collatinginformation from multiple sources —colleagues, family, friends, Web pagesand organisations — about events whichmay or may not happen in multiple loca-tions and timezones, an arbitrary periodahead. Many scheduling applicationsrequire up-to-date knowledge of eachuser’s schedule in order to automate theprocess of finding mutually agreeable

appointments for several attendees.However, the methods used to publishWeb-based schedules (eg train timeta-bles or conference schedules) rendersthem opaque to machine comprehension,thus forcing users to enter these eventsinto their calendars manually.

The Semantic Web supports the creationof Web pages containing structured,meaningful knowledge. These pages can

be parsed to elicit relevant information,and through reference to publishedontologies, agents can understand andreason (using DAML/OWL-baseddescription logics) about this knowledgeto provide user assistance. Thus, bydefining schedules on the SemanticWeb, agents can reason about not onlywhen events occur, but where they occurand which individuals are attendingthem.

The emergence of the Semantic Web has simplified and improved knowledgereuse across the Internet. By committing to various ontologies, an agent can nowunderstand and reason about published information such as calendar events andschedules to meet its user’s needs and provide assistance. We illustrate thebenefits of garnering schedules from the Semantic Web for agent-basedassistance, and introduce other initiatives being pursued by the RDF CalendarTaskforce.

Calendars, Schedules and the Semantic Webby Terry R. Payne and Libby Miller



Access to users’ calendars, includingtheir location and availability, is pivotalfor providing assistance in manydomains. For this reason, various ontolo-gies have been defined for representingschedules, and several agents and appli-cations have been developed to explorehow this resultant knowledge can beutilised. Whilst most calendaring appli-cations use the iCalendar format (RFC2446), a self-contained, text-based repre-sentation suitable for sharing andpresenting event information, the lack ofsemantics limits the level of reasoningthat can be performed. For example,using iCalendar, Jim’s agent cannotreason about whether or not Jim’s friend‘Bob’ is the same ‘Bob’ who ispresenting at a conference. Likewise, itcannot determine if the conference isclose to the user’s current location. Thus,it is unable to make decisions that reflectJim’s preferences, such as schedulinglunch with Bob at a nearby Thai restau-rant.

As part of the RDF Calendar Taskforce,we have developed several tools andvocabularies to address this problem.The development of the hybrid RDFcalendar ontology was based on TimBerners-Lee’s analysis of an RDF adap-tation of iCalendar. Several conferenceschedules have subsequently beenpublished using this ontology (seeFigure 1), and can be queried and viewedusing simple RDF query tools. As RDFsupports the creation of concepts (suchas contact details or locations) asdistributed resources indexed by a URI,schedules can reuse existing conceptsdefined elsewhere. Thus, agents canreason about the resulting schedules(coupled with the ontology definitionsand associated axioms) and henceperform tasks such as the above lunch-scheduling task.

The RCAL agent exploits the structureand semantics of schedules definedusing the hybrid RDF calendar ontologyby providing browsing, importing,reasoning and distributed meetingscheduling functionalities. Users canimport schedules into Microsoft’sOutlook 2000(TM), and compare themwith existing appointments to identifyconflicts. Meetings can be automaticallyscheduled between several users, using

the Contract Net Protocol to locate mutu-ally agreeable times. Notifications asso-ciated with events can be sent to mobiledevices to alert users when they areabout to start. Imported events are alsomonitored, and the user is notified ifthese events change.

Other tools and ontologies have emergedfrom the Calendar Taskforce. A vocabu-lary similar to that used by Palm PDAshas been developed in tandem with toolsthat can synchronise these devices withRDF data. Conversion tools that mapbetween the iCalendar and xCalendarformats (an XML format for iCalendar)have been developed and adapted togenerate the modified Palm format fromthe Mozilla calendar. TranslationServices that convert schedules definedby other ontologies (eg DARPA MeetingAgenda Ontology) into hybrid RDFcalendar markup have been developed;these Web services are located andinvoked automatically when RCalencounters new markup. Members ofthe EC-funded SWAD-Europe projectare experimenting with RSS 1.0 and itsevents module, a simple subset ofiCalendar in RDF. For planningpurposes we require information about

deliverables, meetings, and relatedevents (eg conferences) to be used inscheduling. We use various tools (eg theXHTML-RSS converter) to store, trans-form, merge and view event informationto simplify planning. Our aim is todemonstrate the utility of these tools andthe Semantic Web by ‘eating our owndogfood’, ie using the very tools ourwork is promoting.

Once people make their events informa-tion available in RDF, we can not onlyshow why it is now useful to do this butwe can also use that data to create moreadvanced research scheduling applica-tions. In this sense, calendaring providesa partial answer to the question — howdo we get from here (HTML web) tothere (Semantic Web)?

Link:http://ilrt.org/discovery/2001/04/calendar/

Please contact:Terry R. PayneUniversity of SouthamptonTel: +44 23 8059 3255E-mail: [email protected]

Libby Miller, University of BristolTel: +44 117 928 7113E-mail: [email protected]

Figure 1: A Hybrid RDF schedule containing two events.

<foaf:Person rdf:ID="terrypayne"><foaf:name>Terry Payne</foaf:name><foaf:mbox rdf:resource="mailto:[email protected]"/><foaa:RCalendarAgentName>terry_acm.org-

CalAgent</foaa:RCalendarAgentName></foaf:Person>

<ical:VCALENDAR rdf:id="TAC01"><dc:title>Trading Agent Competition 2001 Workshop</dc:title><dc:contributor rdf:resource="#terrypayne"/><dc:date>2001-10-03</dc:date>

<ical:VEVENT-PROPrdf:resource="http://www.tac.org/2001event.rdf#PainInNEC"/>

<ical:VEVENT-PROP><ical:VEVENT rdf:id="RetsinaTrading">

<ical:DTSTART><ical:DATE-TIME><rdf:value>20011014T134500</rdf:value></ical:DATE-TIME>

</ical:DTSTART><ical:DTEND>

<ical:DATE-TIME><rdf:value>20011014T140000</rdf:value></ical:DATE-TIME></ical:DTEND><ical:LOCATION rdf:resource="#HRTampa" /><ical:ATTENDEE rdf:resource="http://www.daml.ri.cmu.edu/people.rdf#yn" /> <ical:DESCRIPTION>Presentation: Retsina</ical:DESCRIPTION>

</ical:VEVENT></ical:VCALENDAR>



In the next evolutionary step of the Web,termed the Semantic Web, a vast numberof information resources will be madeavailable along with various kinds ofdescriptive information, ie metadata.The Resource Description Framework(RDF) enables the creation andexchange of resource metadata asnormal Web data. To interpret thesemetadata within or across user commu-nities, RDF permits the definition ofappropriate schema vocabularies(RDFS). Managing voluminous RDFdescription bases and schemas withexisting low-level APIs and file-basedimplementations does not ensure fastdeployment and easy maintenance ofreal-scale Semantic Web applications(eg Knowledge Portals and E-Market-places). Still, we would like to benefitfrom database technology in order tosupport declarative access and logicaland physical RDF data independence.This is the main design choice ofRDFSuite, which comprises theValidating RDF Parser (VRP), theSchema-Specific Data Base (RSSDB)and an interpreter for the RDF QueryLanguage (RQL). It was developed in

the context of the IST projects C-Weband MesMuses, both coordinated byINRIA. (http://cweb.inria.fr). The Figuredepicts the architecture of RDFSuite,which is available for download under anOpen Source Software License fromhttp://139.91.183.30:9090/RDF/.

The Validating RDF Parser (VRP) is atool for analysing, validating andprocessing RDF schemas and resourcedescriptions. The Parser syntacticallyanalyses the statements of a givenRDF/XML file according to the RDFModel & Syntax Specification. TheValidator checks whether the statementscontained in RDF schemas and resourcedescriptions satisfy the semanticconstraints derived by the RDF SchemaSpecification (RDFS). Unlike other RDFparsers, VRP is based on standardcompiler generator tools, namely CUPand Jflex, which ensure good perfor-mance when processing large volumesof RDF descriptions. VRP has beensuccessfully used with RDF schemasfrom existing Semantic Web applica-tions, hosted by the Schema Registry athttp://139.91.183.30:9090/RDF/Examples.html.

VRP supports embedded RDF in XMLor HTML documents, XML SchemaData Types and Unicode character set.Users can activate or deactivate semanticconstraints against which validation isperformed and configure the parseraccording to their needs. Anothersubstantial feature is the fetching ofremote namespaces and their integrationin VRP’s internal RDF model. VRPprovides various options for debugging,serialisation under the form of triples orgraphs and complete statistics of vali-dated schemas and resource descriptions.

The RDF Schema-Specific Data Base(RSSDB) is a persistent RDF Store forloading resource descriptions in anobject-relational DBMS by exploitingthe available RDF schema knowledge. Itpreserves the flexibility of RDF inrefining schemas and enriching descrip-tions at any time, and can store resourcedescriptions created according to one ormore associated RDF schemas. Its maindesign goals are the separation of theRDF schema from data information andthe distinction between unary and binaryrelations holding the instances of classes

by Sofia Alexaki, Nikos Athanasis, Vassilis Christophides, Greg Karvounarakis, Aimilia Maganaraki, Dimitris Plexousakis and Karsten Tolle

RDFSuite addresses the need for effective and efficient management of largevolumes of RDF metadata as required by real-scale Semantic Web applications.

The ICS-FORTH RDFSuite: High-level Scalable Tools for the Semantic Web

RDFSuite architecture.



We make an analogy between the corpo-rate memory and the open World WideWeb: both are heterogeneous anddistributed information landscapes andshare the same problem of informationretrieval relevance. In contrast, however,corporate memory has a context, aninfrastructure and a scope limited to theorganisation in question.

As research on the Semantic Web aimsto make the semantic contents of theWeb interpretable by software, wepropose to materialise a corporatememory as a ‘Corporate Semantic Web’,constituted of ontologies, resources (iedocuments or humans) and semanticannotations on these resources (ie on thedocument contents or the personfeatures/competences), where theseannotations rely on the ontologies.

We are studying the main research prob-lems raised by such a ‘CorporateSemantic Web’:• how to detect organisational needs,

and whether to build the CorporateSemantic Web and each of its interre-lated components from several humanexperts or from texts (cf methodologyand architecture)

• how to diffuse and use the memory(information retrieval vs proactivedissemination)

• how to evaluate and evolve thememory (from both designer and userviewpoints).

Ontology and AnnotationConstructionFor the construction of ontologies, theCoMMA method integrates:

• a top-down approach relying on asynthesis of existing ontologies

• a bottom-up approach so as to exploitthe (possibly partly automated) anal-ysis of corporate documents or ofhuman experts’ interviews guided byscenarios

• a middle-out approach relying ongeneralisation and specialisation ofkey concepts of the organisation.

We used this hybrid method to build theO’CoMMA ontology (see below), dedi-cated to organisational memory and toscenarios such as newcomer insertion ortechnological monitoring.

As several corporate information sourcescan help to build an ontology (eg humanexperts, textual or multimedia docu-ments, databases with possibly textual

The ACACIA multidisciplinary team at INRIA-Sophia-Antipolis is working towardsoffering methodological and software support (ie models, methods and tools) forknowledge management (ie for building, managing, distributing and evaluatinga corporate memory) for an organisation or community. This corporate memorycan be materialised in a ‘Corporate Semantic Web’ consisting of ontologies,resources (such as documents or persons) and annotations, possibly withmodelling of multiple viewpoints.

Corporate Semantic Websby Rose Dieng-Kuntz

and properties respectively. The experi-ments we have carried out illustrate thatRSSDB yields considerable perfor-mance gains in query processing andstorage volumes as compared to triple-based RDF Stores.

RSSDB has been implemented on top ofthe PostgreSql ORDBMS. It comprisesLoading and Update modules, bothimplemented in Java. Access to theORDBMS relies on the JDBC interface.Its most distinctive feature is thecustomisation of the database represen-tation according to the employed meta-schemas (RDF/S, DAML-OIL), thepeculiarities of RDF schemas anddescription bases and the target queryfunctionality. The RSSDB Loadersupports incremental loading ofdistributed namespaces by automaticallydetecting changes in already stored RDFschemas or data.

RQL is a typed language, following afunctional approach. RQL relies on aformal graph model that captures theRDF modeling primitives and permitsthe interpretation of superimposedresource descriptions by means of one ormore schemas in various applicationcontexts (recommendations, contentratings, push channels, etc). The noveltyof RQL lies in its ability to seamlesslycombine schema and data querying.

The RQL Interpreter has been imple-mented in C++ on top of PostgreSqlusing a standard client-server architec-ture for Solaris and Linux platforms. Itconsists of three modules: (a) the Parser,which analyses the syntax of queries; (b)the Graph Constructor, which capturesthe semantics of queries in terms oftyping and interdependencies ofinvolved expressions; and (c) theEvaluation Engine, which accesses RDF

descriptions from the underlyingdatabase via SQL3 queries. It supportsXML Schema data types, groupingprimitive aggregate functions and recur-sive traversal of class and property hier-archies. It is easy to couple withcommercial ORDBMSs as well as tointegrate with various Web ApplicationServers. Finally, it provides a genericRDF/XML form of query results that canbe processed by standard XSL/XSLTscripts. An online demo is available athttp://139.91.183.30:8999/RQLdemo/.

Links: http://139.91.183.30:9090/RDF/http://139.91.183.30:8999/RQLdemo/http://cweb.inria.fr/

Please contact:Vassilis Christophides, Dimitris PlexousakisICS-FORTHTel: +30 810 3916 28, 3916 37E-mail: {christop, dp} @ics.forth.gr



data), the SAMOVAR method andsystem rely on partly automatedconstruction and enrichment of ontolo-gies and of annotations from heteroge-neous information sources, by applyingnatural language processing tools ontextual sources. SAMOVAR wasapplied to project memory in the auto-mobile industry.

Ontology and AnnotationRepresentation As several languages can representontologies and annotations in aCorporate Semantic Web, the coopera-tive project ESCRIRE with the EXMOand ORPAILLEUR teams enabled us tocompare three knowledge representation(KR) formalisms (conceptual graphs(CG), object-based KR and descriptionlogics), from the standpoint of the repre-sentation and the handling of documentcontent. Our team focused on the CGformalism.

In order to palliate some limitations ofRDFS for ontology representation, weproposed DRDFS, an extension ofRDFS which enables the expression ofexplicit definitions of concepts and ofrelations, as well as contextual know-ledge. We also proposed the GDLlanguage, a description logic inspired byconceptual graphs.

CORESE Semantic Search EngineRDF, recommended by W3C, allows thecontents of documents to be described inthe form of metadata resting on anontology represented in RDF Schema(RDFS). We developed CORESE(COnceptual REsource Search Engine),an RDF(S) engine based on conceptualgraphs. It enables an RDFS ontology andRDF annotations to be loaded and trans-lated into CG, and then, thanks to the CGprojection operator, allows the base ofannotations to be queried. CORESE alsoreasons on properties of relations(symmetry, transitivity, etc) and onXML-based inference rules, so as tocomplete the annotation base. CORESEwas used in the projects Aprobatiom,CoMMA, ESCRIRE and SAMOVAR,and allows ontology-guided informationretrieval in a ‘Corporate Semantic Web’.

The CoMMA project: Corporate Memory Managementthrough AgentsHeterogeneity and distribution of multi-agent systems can be a solution toheterogeneity and distribution of acorporate memory. In the framework ofthe IST project CoMMA (with ATOS,CSTB, LIRMM, the University of Parmaand T-Nova), we proposed a multi-agentarchitecture for the management of aCorporate Semantic Web, enablingontology-guided information retrieval(the ‘pull’ approach) and proactivedissemination of information to usersaccording to their profiles (the ‘push’

approach). We distinguish agents dedi-cated to ontologies, to documents, tousers and to interconnection betweenagents. The agents are guided by theO’CoMMA ontology, a corporate modeland user models. Some agents havelearning capabilities and can adapt to theusers.

We have also developed algorithms forlearning ontologies incrementally fromRDF annotations on (corporate) Webresources. This method can be used tobuild classes of documents automaticallyfrom their annotations.

Figure 1: CORESE semantic search engine interface.

Figure 2: Corporate Memory Managements through Agents — CoMMA architecture.



The Cybersecurity team of the EC’sJoint Research Centre (JRC) in Ispra,Italy has recently completed work on afully compliant implementation of theW3C’s new privacy protocol, P3P,which uses XML files for the automatedexchange and interpretation of websiteprivacy policies and their matchingagainst user preferences.

In the next stage of our research, the JRCwill use this base platform to investigatehow Semantic Web (SW) technologycan improve this system. The project hasbeen divided into a series of distinctstages.

The first stage is already under way andconsists of the development and agree-ment of an ontology for data protection.

One of the most crucial improvementsthat the SW can make to the P3P plat-form will be to provide an interoperable,machine interpretable ontology for dataprotection. This will provide thefollowing benefits:

• a common, interoperable vocabulary,which reduces misinterpretation ofbasic principles by technologists andlegal experts and enhances interoper-ability between different systems

• clear understanding of terms allowingease of translation between alternativeontologies

• clear separation of vocabulary andsyntax meaning that the same vocabu-lary can be plugged into different dataprotection systems.

• a clearly documented developmentprocess offering clarity, authority andcommon agreement on terms by alarge group of stakeholders.

An agreed vocabulary should be thebasis for any privacy system. Significantproblems were found in P3P’s firstversion, mainly arising from inadequa-cies in its expression of data protectionconcepts such as those found in regula-tory frameworks (eg the EU data protec-tion directive and the OECD guidelines).

The JRC is developing a consensusprocess for capturing the knowledge ofdomain experts such as data commis-sioners. This is in conjunction withresearchers from the University ofAberdeen, whose prototype ontologycapture process will be used. Thisprocess includes input from psycholo-gists and conceptual modeling researchand has been put through several testcases.

Despite incompleteness in certainrespects, P3P provides a sound founda-tion for such an ontology. Although P3Pis not expressed in standardised ontologysyntax, it represents, through the W3Cprocesses that have underpinned it, a fiveyear consensus process for a data protec-tion vocabulary. Given a formally docu-mented consensus process and improve-ments in the data typing schema, andpurpose and recipient taxonomies, theexisting version of P3P provides a veryuseful starting point.

by Giles Hogben and Marc Wilikens

The Joint Research Centre of the EC is building an experimental privacy protectionagent using Semantic Web technology and the Worldwide Web Consortium’s P3Pprotocol. The agent automates the process of protecting a user’s privacy throughparsing and comparing privacy policies and user preference sets.

Joint Research Centre of the EC to use Semantic Web for New Privacy Agent

Two scenarios were studied: the inser-tion of new employees and technologicalmonitoring. The CoMMA system,implemented in JAVA on the FIPA-compliant multi-agent platform JADE,integrates our search engine CORESE.

Applications and Further WorkOur work on Corporate Semantic Webswas applied to project memory, techno-logical monitoring, knowledge servers inthe automobile industry (Renault), in theconstruction sector (CSTB), in telecom-munications (T-Nova, TILAB) and inthe biomedical domain. We also takepart in the IST thematic networkOntoWeb (Ontology-Based Information

Exchange for Knowledge Managementand Electronic Commerce).

In further work we intend to study acorporate memory distributed amongseveral cooperating companies orcommunities. This memory will bematerialised through a Semantic Weboperating between organisations orcommunities, possibly handling multipleontologies and multiple viewpoints andpossibly requiring Web mining, forscenarios including project memory,technological monitoring and skillsmanagement.

The work described above was carriedout by Olivier Corby, Alexandre Delteil,Rose Dieng-Kuntz, Catherine Faron-Zucker, Fabien Gandon, Alain Giboin,Joanna Golebiowska and CarolinaMedina-Ramirez.

Links:http://www.inria.fr/Equipes/ACACIA-eng.htmlhttp://www.inria.fr/acacia/Publications

Please contact: Rose Dieng-Kuntz, INRIATel: +33 4 92 38 78 10E-mail: [email protected]



The second stage will be to ensure thatthe proposed extensions are backwardcompatible. For this purpose, an XSLTstylesheet will be created to perform atranslation between the two versions.The concepts of the ontology will also beplaced within a clear W3C style specifi-cation.

The third stage will be to develop a rulesystem, which agents can use to makedecisions based on the new framework.The rules, as in the previous system, lookfor a set of conditions in an RDF privacypolicy, and the matcher executes abehaviour depending on the conditionsatisfied. The condition is identifiedusing an RDF query language. The rule-matching process proceeds as follows:• retrieve P3P privacy policy and load

set of rules• attempt to match conditions from each

rule in turn• when first rule matches, perform

behaviour specified by that rule• no rules fired is an error (therefore a

catch-all rule must be included).

At this point, there will be a workingsystem, as in the Figure.

The fourth stage will be to refine the rulesystem and policy language using testcases from European law. In the firstphase of the JRC’s P3P implementation,work was done on expressing theEuropean Data Protection Directivesusing P3P vocabulary. This revealed thatthere are some inadequacies in the

vocabulary, particularly when describingwhether data is given to recipientsoutside of the EU and in the descriptionof the purposes of data collection. Theforthcoming SW version will, it ishoped, have sufficient conceptual flexi-bility to express the European DataProtection Directives. In any case, thesewill act as a litmus test for the system.

Finally, the research will cover ways ofusing the SW to make privacy informa-tion more accessible to other systems. Atpresent, P3P functions only within thecontext of http transactions. The use ofSW technology, as a standardised know-ledge transfer medium, may help otherareas of technology to use P3P.Examples might include ubiquitouscomputing, smart card technologies andIRC chat rooms. These developmentsput P3P in a good position as SW tech-nology becomes more widespread. Asknowledge sharing moves increasinglytowards the use of the SW, the existenceof an integrated privacy framework willfacilitate association of privacy prefer-ences and policies with data as theymove between heterogeneous sources.

Links: More information and a demonstration of our current platform: http://p3p.jrc.it

W3C’s P3P Privacy protocol home page:http://www.w3.org/P3P

RDF schema for P3P:http://www.w3.org/TR/p3p-rdfschema/

OWL Web ontology home page:http://www.w3.org/2001/sw/WebOnt/

Please contact:Giles Hogben, European Commission Joint Research Center, ItalyTel: +39 033 278 9187E-mail: [email protected]

Semantic-Web-enabledprivacy agent.



Approaching semantically definedknowledge on the Web is not the end ofinteroperability problems. Indeed,knowledge providers use differentlanguages, axiomatic or conventions forexpressing knowledge. Importing know-ledge in the context of different applica-tions will require adapting it to the targetcontext (language, axiomatic or conven-tion). This adaptation can be processed apriori or on the fly and the transforma-tion involved can be automatically ormanually generated.

The action of these transformationsranges from minor syntactic formattingto complex processing such as deducingconsequences and generalising them. Itis important for applications to be confi-dent of the properties of these transfor-mations. These properties can be relatedto the syntax (eg correct expression ofthe result in a particular language), thesemantics (eg preservation of the mean-ings of knowledge by the transforma-tion) or the pragmatics (eg preservationof the reader’s interpretation by transfor-mation). They may also concern thetransformation processing (eg correcttermination) or its results (eg loss ofinformation in the result).

One of the goals of the Exmo action is toassess the properties satisfied by trans-formations. This objective requires: • the capacity to analyse transforma-

tions into units assembled by preciseconstructors

• the knowledge of properties that cancharacterize transformations and theirbehaviour with regard to the construc-tors

• the opportunity to prove propertiesfrom elementary transformations.

Exmo deals with these three problemsand aims to build an environment inwhich users can compose transforma-tions and formally characterise theirproperties.

In collaboration with the Fluxmediacompany we developed theTransmorpher environment that allowsthe composition of XML transforma-tions. The unit transformations are eitherexpressed in XSLT (in which case theproperties are asserted) or in a simplifiedrule transformation language (fromwhich properties can be assessed). Thesetransformations can be composed insubroutine, sequence, parallel or loopswhich constitute the main constructorsof the system. The environment isportable and freely available (seeFigure).

The transformations developed withinTransmorpher will have to be annotated

by the properties that they satisfy. Inparticular we are investigating semanticproperties (tied to the model theoreticsemantics of the languages involved)and their relations (eg interpretation-preservation implies consistency-preser-vation). These properties have beenapplied to an XML encoding of descrip-tion logics and this effort should be reit-erated in the next W3C-proposedlanguages.

The next step consists in providingTransmorpher users with tools forproving the properties of a particulartransformation. Our current efforts usethe relations between properties in orderto establish the minimal propertiesenjoyed by the composition ofTransmorpher constructors. We wouldalso like to experiment with proofcheckers for the inductive proofs oftenused for proving the equivalence oflanguages.

by Jérôme Euzenat

Sharing knowledge on the Semantic Web will require the transformation andadaption of knowledge to client needs. The complexity of the representations tobe transformed can hide the effects of transformations, though the use ofknowledge dictates them. Exmo aims to provide tools for characterising the actionof these transformations in syntactic, semantic or semiotic terms.

Exmo: Assessing Semantic Properties of Knowledge Transformations

Screenshot of the Transmorpher graphic environment.



The Semantic Web vision of merginginformation from a diversity of Internetsources presupposes methods forprocessing diverse types of descriptivemetadata. The past decade has seennumerous initiatives to standardise meta-data semantics for particular applicationcommunities, from multimedia andcommerce to scholarly publishing. Inpractice, there is some overlap betweenthese communities of practice, andimplementors often draw pragmaticallyon available standards in designingmetadata for particular applications,‘mix-and-matching’ as needed.

Processing metadata scalably willrequire infrastructures to interpret, trans-late, and convert among various meta-data languages automatically. Theconstruction of such infrastructures willhowever require the availability of meta-data dictionaries in forms usable by soft-ware agents, middleware applications,and indeed human beings. Such dictio-naries should ideally be based onconventions, shared across metadatacommunities, for declaring metadatavocabularies machine-understandably.

The objective of the CORES Project, anAccompanying Measure under theSemantic Web action line of the EU’sFifth Framework Programme, is to facil-itate the sharing of metadata semanticson several levels. At the level ofstandardisation initiatives, the project isbringing together key figures from majorstandardisation activities in a StandardsInteroperability Forum to discuss princi-ples and practicalities of ‘playingtogether’ in the Web environment. Theparticipants include representatives frominitiatives as diverse as DCMI, OASIS,DOI, CERIF, MPEG-7, IEEE/LOM,GILS, ONIX, and the W3C SemanticWeb Activity (with apologies for theunexpanded acronyms). While under-taken with reference to Semantic Webtechnologies, the Forum includescommunities that are pursuing differentstrategies for interoperability.

At the level of infrastructure, CORES isworking with organisations that aredesigning or building value-addedservices for metadata-using applications.Building on the prior SCHEMAS Project(2000-2002), CORES is expanding aregistry of application profiles andliaising with related activities on devel-

oping good practice for interoperabilitybetween specialised registries. Thiseffort will include a hands-on ‘profile-writing’ workshop with implementors.

One goal is to test the SCHEMAS modelfor declaring application profiles assimple sets of RDF assertions — a formwhich in principle lends itself to merginginto ‘semantic landscapes’ of informa-tion providers.

Project partners are Pricewaterhouse-Coopers (Luxembourg), UKOLN,University of Bath (UK), Fraunhofer-Gesellschaft, and SZTAKI.

Link: http://www.cores-eu.net/

Please contact: Thomas Baker, Fraunhofer-Gesellschaft Tel: +49 30 8109 9027E-mail: [email protected]

by Tom Baker

Can different standards communities agree on principles to help translate amonga diversity of metadata languages?

CORES — A Forum on Shared Metadata Vocabularies

In the context of the Web, and of theSemantic Web infrastructure, transfor-mations could be made available formany other parties that might want tointegrate them within their ownprocesses. In order to establish the prop-erties satisfied by the resulting transfor-mation, it will be necessary to be able toassess those of the gathered transforma-tions. For that reason, we want to extendthe system presented above with thecapability of importing and exportingtransformations, their properties and, if

available, the proofs of these properties.The importers would then be able tocheck the proofs and thus trust the prop-erties of the imported transformations.This is an instantiation of ‘proof-carrying code’ to transformations.

Besides classical properties tied to thesyntax and semantics of the representa-tions, the overall goal of Exmo is to beable to take into account properties tiedto their pragmatics, such as the rhetoricof a discourse or the semiotic interpreta-

tions of readers. For instance, this wouldenable us to establish that the sameimpact is provided by a multimediapresentation before and after its transfor-mation. This requires investigations intothe field of semiotics.

Links: http://www.inrialpes.fr/exmohttp://transmorpher.inrialpes.fr

Please contact:Jérôme Euzenat, INRIATel: +33 476 61 53 66E-mail: [email protected]



The success of the Internet and XMLcreated a mistaken belief in a global B2Bmarket where businesses would be ableto communicate with each other withoutany prior agreements. These technolo-gies seemed to overcome the mostsignificant obstacles of traditional elec-tronic data interchange (EDI) standards.After the initial excitement however,people realised that while XML providesa means to exchange data between appli-cations, it does not guarantee interoper-ability. While XML provides a syntaxthat can be used for data transfer in B2B,it does not provide semantics, since tagshave no predefined meaning. Themeaning of XML vocabularies is definedby the document designer and is speci-fied outside of XML itself. This resultedin a proliferation of XML-based e-busi-ness vocabularies within the first few

years of XML’s release. Havingcontributed to the work of UN/CEFACTin the Techniques and MethodologyWorking Group (TMWG) and ebXML,as well as conducting related researchprojects such as A-XML and Harmonise,we feel that a true B2B solution needsmuch more than a purely syntax-basedapproach. Here we elaborate on the openissues that we see on the road towardssyntax-independent B2B.

In the future, we expect some vocabu-laries to disappear and others to merge.Nevertheless, a certain number of well-known ‘standard’ vocabularies willcoexist. This means that business part-ners must either agree on a certain e-business vocabulary or map betweentheir mutual vocabularies. This is notpracticable with a large number of busi-

ness partners. Consequently, a commonsyntax-neutral ontology capturing thebusiness semantics independently ofvocabulary is desirable. Furthermore, aclearly defined representation of thebusiness semantics in each supported e-business vocabulary is necessary. Thisapproach minimises the effort requiredsince the number of mappings can bekept as low as the number of vocabu-laries. In this context, two key aspectsare highly relevant for the researchcommunity, namely, how to build theontology and the mapping between theontology and the business vocabularies.

A major failure of today’s B2Bapproaches is the issue of overloadeddocument types. This is the result of thedesire to create one single purchaseorder, one single invoice etc, for thewhole world. In order to meet therequirements of any company in anybusiness sector in any geographicalregion, document types are created bycollecting (and structuring) all dataelements that might be relevant for anycompany in any sector in any region. Asa consequence, only a very smallpercentage (usually less than 5%) of dataelement types are instantiated in a docu-ment. To guarantee interoperability,partners have to agree on the relevantsubset of the document type prior to theexchange. Thus, documents must becompliant not only with the documenttypes (called valid documents in XML)but also with industry-, region- orpartner-specific agreements.Accordingly, we identify a third keyaspect for the research community: theautomatic adjustment of general docu-ment types according to given contextparameters such as industry, region, andbusiness process.

When it was introduced, XML seemed to promise a solution to traditional B2Bproblems. Today, however, we are faced with a multitude of different XML-basede-business vocabularies that are not interoperable. Even in the case of businesspartners supporting the same e-business vocabulary, interoperability is notguaranteed since they may be implementing different subsets of the samevocabulary’s document type. A research team at the University of Viennaaddresses these problems with a framework that aims to align concepts knownfrom the Semantic Web and the ebXML initiative.

Towards Syntax-Independent B2Bby Birgit Hofreiter, Christian Huemer and Werner Winiwarter

B2B Framework based on Ontologies and the Semantic Web.



In order to address these three aspects wesuggest drawing on approaches from theSemantic Web and electronic businessXML (ebXML). In the Semantic Webcommunity, languages such as RDF andDAML have evolved that can beemployed to develop the syntax-neutralontology. Furthermore, there is ongoingresearch concerning the transformationof an ontology’s semantics into XMLrepresentations of certain e-businessvocabularies.

EbXML core components are orientedalong the most extensive knowledgebase in the business world, the trade dataelement dictionary UN/TDED.Therefore, ebXML core components,which will also be supported byBoleroNet, EAN/UCC, Open-Applications Group (OAG), OpenTravel Alliance (OTA) and SWIFT,provide the best semantic foundation fora global business ontology. As ebXMLalso allows the specification of businessprocesses, their activities and theiractivity-specific business data, this facil-itates the adjustment of business docu-

ments to the needs of a business processon the basis of core components, ratherthan through the use of generic docu-ment types. Thus, ebXML provides agood starting point for our third identi-fied aspect.

By aligning ontology and ebXMLconcepts we propose the frameworkdepicted in the Figure. This frameworkis based on four major steps. In the firststep the semantics of ebXML corecomponents are defined in an RDFS-based ontology. The second step definesmappings between the ontology andRDF representations of e-businessvocabularies, as well as between thelatter RDF representations and the DTDsor XML schemas of the e-businessvocabularies. Beyond traditional onto-logical approaches, we take on theebXML idea to further restrict a generaldocument ontology to the specific needsof a given business activity. This refine-ment will lead to a certain view of thegeneral document ontology and must bespecified by means of a constraintlanguage. Steps one to three must be

done manually and are depicted in thewhite triangle, whereas the last stepdepicted in the grey triangle will bederived automatically. Using the infor-mation from the previous steps, a morerestrictive and appropriated XMLschema or DTD for a specific activitywill be created. The XSDs and DTDs area subset of their corresponding generalones. Further restrictions expressed in adeclarative language will accompany thedocument types. This frameworkfocuses on a mapping between differente-business vocabularies and at the sametime guarantees their semantic equiva-lence in support of a specific activity.

Links:ebXML: Homepage of the ebXML Initiative: http://www.ebxml.orgSemantic Web Portal: http://www.semanticweb.org

Please contact:Birgit Hofreiter, University of ViennaTel: +43 1 4277 38434E-mail: [email protected]

One way of creating a taxonomy for aknowledge domain is by identifying anumber of different aspects or facets ofthe domain and then designing onetaxonomy for each facet. Several studiesin information, library and cognitivescience have shown that in almost everyknowledge domain we can indeed distin-guish a number of facets, or planes ofunderstanding. A faceted taxonomy isactually a set of taxonomies, calledfacets, each of which is a set of termsstructured by a specialisation/generalisa-tion relation. Using a faceted taxonomy,the indexing of objects is done by associ-ating each object with a compound term,ie with a combination of terms coming

from different facets. It was recognisedlong ago that a faceted taxonomy hasseveral advantages over a single hierar-chical taxonomy, including conceptualclarity, compactness and scalability. Forexample, consider two schemes forindexing the objects of a domain, thefirst using a single taxonomy consistingof 100 terms, and the second using afaceted taxonomy consisting of 10 facetseach having 10 terms. The first schemehas 100 indexing terms while the secondhas 1010 (10 billion) compound indexingterms! Although both schemes have thesame storage requirements, ie each onerequires storing 100 terms, the second

has tremendously more indexing termsthan the first.

However, faceted taxonomies have amajor drawback that prevents theirdeployment and use for real and large-scale applications like the Web. Thisdrawback comes from the fact that it ispossible to form a large number ofinvalid compound terms, that is, combi-nations of terms that do not apply to anyobject of the underlying domain. Forexample, refer to the faceted taxonomyfor a tourist information applicationshown in Figure 1 and consider the termsWinterSports from the facet Sports, andCrete from the facet Location. The

by Yannis Tzitzikas, Nicolas Spyratos, Panos Constantopoulos and Anastasia Analyti

Which would be easier to remember: one thousand individual terms or threefacets of ten terms each?

Extended Faceted Taxonomies for Web Catalogs



compound term WinterSports.Crete isinvalid, as there is never enough snow inCrete! In contrast, the compound termSeaSports.Crete is certainly valid. Theinability to infer the valid compoundterms may create problems in objectindexing (laborious and/or erroneousindexing), and in browsing (an invalidcompound term will yield no objects).Due to such problems, existing Webcatalogues have a strictly hierarchicalstructure like the one shown in Figure 2.Such taxonomies suffer from severalproblems such as incomplete termi-nology, huge size (for example thetaxonomy of Open Directory consists of300 000 terms) and confusing structure.

Being able to infer the validity of acompound term in a faceted taxonomywould be very useful for facili-tating indexing and forpreventing errors, especially incases where indexing is done bymany human editors (indexers).For example, Web pages in theOpen Directory are currentlyindexed by more than 20 000volunteers. Moreover, if wecould infer the valid compoundterms in a faceted taxonomy thenwe would be able to generatenavigation trees (like thetaxonomies of Web catalogues)on the fly, consisting of nodesthat correspond to validcompound terms. However,manually defining the set of validcompound terms even for facetsof relatively small size would bea formidable task for the

designer. For example, Figure 1 showsthe partition of compound terms into setsof valid and invalid terms.

To alleviate this problem, we havedefined two extensions of facetedtaxonomies, which we call PEFT andNEFT, which allow the specification ofvalid compound terms in a flexible andefficient manner. A PEFT (PositiveExtended Faceted Taxonomy) is afaceted taxonomy enriched with a set Pof compound terms known to be valid,while a NEFT (Negative ExtendedFaceted Taxonomy) is a facetedtaxonomy enriched with a set N ofcompound terms known to be invalid.The designer simply declares a small setof valid or invalid compound terms, andother (valid or invalid) compound terms

are then inferred by a mechanism basedon semantic implication. Figure 3 showshow we can specify the valid compoundterms of the faceted taxonomy ofFigure 1, ie the sets ‘Valid CompoundTerms’ and ‘Invalid Compound Terms’as enumerated in that figure, byemploying either a PEFT or a NEFT. Ineach case we can dynamically derive anavigation tree such as the one shown inFigure 2 that can be exploited duringobject indexing and browsing.

Our approach can be used for developingWeb catalogues that offer completenavigation trees, require less storagespace, and are more comprehensive andscalable. Furthermore, taxonomiesdesigned according to our approach canbe integrated or articulated more easily

than hierarchical ones. Furtherresearch includes extending thisapproach to define an algebraicsystem for taxonomies with oper-ators that allow the specificationof valid compound terms usingboth positive and negative sets ofcompound terms.

This work is the result of basicresearch conducted at ICS-FORTH over the last two years.

Please contact:Yannis Tzitzikas, ICS-FORTHTel: +30 810 391 623E-mail: [email protected]://www.csi.forth.gr/~tzitzik

Figure 3: Two extended faceted taxonomies.

Figure 1: A faceted taxonomy consisting of two facets, Sportsand Location, and the partition of the set of compound termsto the set of valid and the set of invalid terms.

Figure 2: A hierarchical organisation of the valid compoundterms of the faceted taxonomy of Figure 1.



W3C is developing standards for therepresentation of ontologies to constrainthe vocabularies of resource descriptionsbased on RDF. Such ontologies willallow distributed authoritative definitionof vocabularies that support cross-refer-encing. Such ontology representationsare planned to fulfil the role currentlyundertaken by thesauri. Therefore amigration path is required from currentthesauri to ontologies, or support fortheir co-existence if those ontologies areto be adopted and assimilated intoexisting information retrieval infrastruc-ture.

The structure of thesauri is controlled byinternational standards that are amongthe most influential ever developed forthe library and information field. Themain three standards define the relationsto be used between terms in monolingual

thesauri (ISO 2788:1986), the additionalrelations for multilingual thesauri (ISO5964:1985), and methods for examiningdocuments, determining their subjects,and selecting index terms (ISO5963:1985). The general principles inISO 2788 are considered language- andculture-independent. As a result, ISO5964:1985 refers to ISO 2788 and uses itas a point of departure for dealing withthe specific requirements that emergewhen a single thesaurus attempts toexpress ‘conceptual equivalencies’among terms selected from more thanone natural language.

The ISO standards for thesauri (ISO2788 and ISO 5964:1985) are developedand maintained by the InternationalOrganization for Standardization,Technical Committee 46 whose remit isInformation and Documentation — not

IT. ISO 5964:1985 is currently under-going review by ISO TC46/SC 9, and itis expected that among changes to it willbe the inclusion of a standard inter-change format for thesauri. To facilitatethe growth of the Semantic Web, itwould be sensible to try to ensure thatsuch an interchange format is as compat-ible with Semantic Web ontology repre-sentations as possible.

Several proposals have arisen forthesauri interchange formats based oneither RDF or DAML+Oil. The majorproblems with these is that either theycannot accommodate the multiple inheri-tance common in many multilingualthesauri or that the semantics of thesauriin the ISO standards are not as precise asthese languages require. The links inthesauri hierarchies define the top termin the hierarchy, and the broader or

If Semantic Web technologies using RDF are going to be adopted and assimilatedas HTML and XML have been, a clear migration path from present technologiesto new ones is required. Thesauri are used throughout the information retrievalworld as a method of providing controlled vocabularies for indexing and querying.

Migrating Thesauri to the Semantic Webby Michael Wilson and Brian Matthews

The thesaurus model in UML.



Documents on the Web are often deeplystructured, and this can be the origin ofincompatibility between different viewsof the same object. Links are the essenceof hypertext, but they are meaningfulonly if their semantics is clear, andperceivable by the user. As a conse-quence, it can be useful to have lightlystructured documents, where someelements can be seen as ‘semantic items’that identify concepts chracterising thespecific parts of the document. Links canshare basic semantic categories withthese semantic items, and can be imple-mented in the XLink framework. Thisquite simple approach leads to a consid-erable enhancement of navigation possi-bilities, and makes it possible to presentdocuments apropriately for a real adap-tive hypertext environment.

The Hyperlink Association ModelWhen reading, our attention is oftencaptured by words (anchors) that leadour mind to other documents. In the Web

context, documents, whatever theirorigin, are seen as resources. We canmodel the association process in thefollowing way:• the anchor leads to a concept• the concept is related to other concepts• the new concept is related to some

resources.

This basic association mechanism (seeFigure 1) is totally independent of thedocument structuring. In the data space,documents are connected by extensionallinks. In the concept space — a simpli-fied version of the Semantic Web archi-tecture’s ontology level — associationsamong concepts implement intensionallinks among documents. Two questionsnow arise:• how can we implement the link from

resources to concepts• how can concepts be linked together.

A simple and effective way to implementintensional links is to identify the

semantic items. This can help in severalcases (for example, ‘The Frenchemperor’ can be an implicit reference to‘Napoleon’). We can also characteriseeach semantic item with a specificsemantic category (eg person, location,date, taxonomy) in order to tailor a docu-ment to specific user interests, eg areader interested in space-time associa-tions will get location and date itemsemphasised.

The second question directly leads to theinteraction metaphor issue. In addition tothe case of taxonomic classifications,where we can make use of well-knownthesaurus techniques, space and time canfunction as very powerful associationmechanisms. A semantic item can pointto a location, then, using an interactionmetaphor based upon space, the user caneither jump to other resources linked tothe same location, or select a differentlocation, and then find other resourcesrelated to this new location. This simple

Semantic characterisation can considerably enhance navigation possibilities andmakes possible the presentation of documents for a real adaptive hypertextenvironment.

Semantic Characterisation of Links and Documentsby Silvia Martelli and Oreste Signore

narrower coverage of terms down thehierarchy. There are also links betweenhierarchies to show equivalence indifferent languages, or similar meaningin the same language.

However, the hierarchical links inthesauri are semantically overloaded,and the potential exists using SemanticWeb ontology representations to developontologies with less overloading. Theterms ‘broader’, ‘narrower’, ‘used for’,‘related’ and ‘equivalent’ are not definedby precise semantics. Therefore theproposals are either too precise to becompatible with some existing thesaurior include explicit statements of seman-tics which are seen to be unacceptable toother thesaurus developers and users.

We have developed a proposal for athesaurus interchange format expressedin RDF to overcome these limitations,which has been applied to one largemulti-lingual thesaurus for evaluation byusers - ELSST. It is planned to representmany more thesauri in this representa-tion, and to show how they can both bemigrated into Semantic Web ontologies,and how such ontologies allow thethesauri from different domains to berelated to each other.

ELSST: a multilingual thesaurus(English, French, Spanish, German) forthe social science domain has beenrepresented in the Thesaurus InterchangeFormat. The thesaurus has beenproduced containing 49 hierarchies,

incorporating 1456 preferred terms.Following the initial developmentfurther translations of terms into Finnish,Norwegian, Danish & Greek areplanned, as are the inclusion of termsrelated through inexact translations inaddition to the exact translations alreadyincluded. Also the CESSDA group ofEuropean Data Archives has agreed toadopt ELSST as the EuropeanControlled Vocabulary for SocialScience.

Link: http://www.limber.rl.ac.uk/External/SW_conf_thes_paper.htm

Please contact:Michael Wilson, Brian Matthews, CLRCTel: +44 1235 44 6619E-mail: [email protected]



hyperlink association model can beimplemented through a document, linkand user model.

Semantic Model of DocumentsIn XML documents we must clearlydistinguish between structural andsemantic information, which can beassociated with elements or parts ofthem. For the sake of simplicity andeffectiveness, it is possible to define alimited and rough general set of semanticcategories, that can be structured in athesaurus-like fashion. These categoriescan be shared by a wide variety of users,can be used to define a user profile and

can semantically characterise variousparts of the documents and links. We canalso specify a weight, stating the rele-vance of the concept in the documentcontext.

Link Taxonomy and ModelLinks allow navigation on the Web, andcan implement the abstraction mecha-nisms needed to move from data space toconcept space. Semantic qualificationsof links explicitly identify their meaningin the document and the role of involvedresources. The reason why the link hasbeen inserted in the document, ie thenature of association (geographical,

explicative, etc), can be explicatedthrough the link’s semantic type.Different types of links can suggestdifferent and specialised interactionparadigms (time, map, classification,etc) having an enormous effect on thepotential association mechanism: twodocuments can be linked through anintensional link existing in the conceptspace, even without the extensional linkbeing specified in the document.

User ModelAs a first approximation level, the usermodel is defined in terms of an essen-tially dynamic profile, tightly related tothe semantic model of documents andlinks. The user profile is defined in termsof semantic categories, link types andlink roles. For each of these, a degree ofinterest (weight) is stated.

A Simple ExampleTake the following fragment of an XMLdocument, containing some semanticitems:

In <abwr:si st=”date”> 1812 </abwr:si><abwr:si st=”person” canonicalName =“Napoleon”> the French emperor</abwr:si> invaded Russia ...

This document may have been entered asit is, or may be the result of a morecomplex process, involving a databasesearch and data processing.

A software agent can take this documentas input, producing a richer XML docu-ment, where the expression “the Frenchemperor” becomes the anchor of anextended link (using XLink termi-nology).

When producing this document and alsowhen displaying it, the software agentcan examine the user profile, in order toproduce a personalised document thatwill be displayed by the browser, asshown in Figure 2.

Please contact:Oreste Signore, ISTI-CNRTel: +39 050 3152995E-mail: [email protected]

Figure 2: Document with anactivated semanticitem.

Figure 1: Document and

Concept Space.



Knowledge-intensive organisations aremore common than you might expect:R&D departments and education centresare among those which first come tomind, but any community of interest, anyteam, any collaborative group that main-tains a large amount of distributed andsemi-structured data can fit in this cate-gory.

As Intranet solutions have increased inpopularity, they have become populatedwith documents, forms, calendars ofevents, news, link collections and appli-cations such as databases, newslettersand forums. As a consequence, mostorganisations have ended up with a hugerepository of semi-structured knowledgedistributed over the Intranet. This hasresulted in some common maintenanceand accessibility problems. On the onehand, the administration of these know-ledge sources cannot be centralised,because they are maintained only as longas they are useful for those who writethem. On the other hand, all members ofthe organisation, regardless of differ-ences in cultural background and level ofexpertise, should be able to find whatthey need: the form for initiating anadministrative procedure, the applicationto broadcast urgent news to a communityof interest, someone with a specific set ofskills to whom a task can be assigned, atailored course to learn how to performan unfamiliar operation, etc.

So far, the combination of Web tech-nologies (XML, J2EE, SOAP, etc) andknowledge management methodologieshas proved able to address some of theseproblems. Intranet portals, based on acollaborative and distributed editorialmodel, can provide a single point ofaccess and a good content managementsolution. Moreover, the development of

component-based applications, in partic-ular with emerging Web services tech-nology, provides the strong decouplingand easier composition required by thesekinds of data-intensive Web applica-tions. However, what our state-of-the-arttechnology can provide is good humanaccess to this knowledge. While a goodIntranet search engine can lead the userto a set of knowledge sources, because ofits lack of common sense, it is unable toanalyse them.

Semantic Web technology, with theexplicit introduction of a formal seman-tics for each knowledge source, willmake it possible for a machine to auto-matically process the source and in away, to understand it. In particular, theuse of ontologies, which are explicitconceptualisations of a shared under-standing of a domain, can enable areasoning service to analyse the know-ledge stored in the sources. In addressingthe common problems which arise in theIntranet environment of a typical know-ledge-intensive organisation, the e-Service Technologies unit ofCEFRIEL is working on two projectsand will commence a number of furtherprojects in the coming autumn.

In collaboration with the KM compe-tence centre of Getronics, we arecurrently running the sKM (semanticKnowledge Management) project. Thegoal of sKM is to use ontologies as thecore component of a methodology toselect and configure the most suitableKM solution for a specific organisation.In order to achieve this, we validate thepossibility of integrating into a frame-work a set of freely available tools fordesigning and maintaining ontologies.We monitor the available languages andtools and, so far, we have selected

DAML+OIL as the language, OILedv3.4 as the editor and FaCT v2.2 as thereasoner. Moreover, we have conceivedand implemented two tools: a graphicvisualiser VisualDAML+OIL and an adhoc report generator, ReportDAML+OIL.

We use ReportDAML+OIL to producethe information needed by a technicianto configure a KM solution. The nextstep is to study the possibility of inte-grating current state-of-the-art KM solu-tions with reasoner-able services in orderto address the common maintenance andaccessibility problems. The options weare considering include a framework forautomatic publication of content basedon the semantics associated with thecontent, a mechanism to support theannotation which deduces implicitknowledge from that which is explicitlygiven, and a solution to empower thebasic syntactic search mechanism widelyavailable with the use of the givensemantics. This means the search enginecould learn to find what you mean, ratherthan simply what you say!

In the meantime, we’re implementingCOPPER (CEFRIEL Open PortalProject for Enterprise Resources). This isan industrial-strength portal solutionbuilt only with open-source technolo-gies (Apache, Tomcat, Velocity,PostgreSQL, Java). We have alreadyused it for building an Intranet portal andwe plan to adopt it as our own testbedand training field for the prototypalimplementation of those reasoner-ableservices described above. In the comingautumn we will start several newprojects. In ‘Semantic Enhanced e-Learning’, we will study the problemswhich arise in populating and main-taining an open learning object reposi-tory with special attention paid to the

by Emanuele Della Valle and Maurizio Brioschi

Knowing what you know is increasingly a real problem in many organisationswhose core competence is knowledge. We believe that using state-of-the-art Web-based Knowledge Management (KM) technologies will not be sufficient in theimmediate future, since the lack of formal semantics will make it extremely difficultto make the best use of the massive amount of stored data.

An Ontology-Oriented Solution for Knowledge-Intensive Organisation



construction of ad hoc courses.Moreover, we will start a technologytransfer project that concerns both theskill management and the accessibilityof a CAD-drawing repository. Last butnot least, we’re looking at the possibilityof using semantic-enabled Web services

in an e-government Intranet applicationintegration scenario.

People involved: Maurizio Brioschi,Stefano Campanini, Paolo Castagna,Emanuele Della Valle, Marco Riva,Nicola Simeoni.

Links: http://etechdemo.cefriel.it/swhttp://www.cefriel.it/etech

Please contact:Emanuele Della ValleCEFRIEL, Politecnico di Milano, ItalyTel: +39 02 23954 324E-mail: [email protected]

In order to take advantage of the hugeamount of information available on theWeb and in document databases, it isnecessary to design efficient techniquesfor retrieving, extracting and queryingdocuments. Ontologies are playing anincreasing role in these tasks, especiallyfor content-based annotation and manip-ulation of documents. The objective ofthe ESCRIRE project is to use anontology to annotate a set of abstracts ofbiological documents extracted from theNIH Medline public database, and thento query the annotated documents withina knowledge representation (KR)formalism. In the following, we restrictour attention to the use of a descriptionlogic, namely the RACER system,within the ESCRIRE project.

In order to manipulate documents bytheir content, annotations are attached todocuments and a domain ontology hasbeen designed for this purpose. Theannotations and the ontology aredescribed within a pivot language basedon XML. This pivot language relies on aset of syntactic rules controlled by aDTD. The pivot language plays the roleof a bridge between documents and thedescription logics (DL) formalism: everyelement in the ontology and every anno-tation have a corresponding element inthe DL formalism. The pivot language isalso used to describe SQL-like queries,which are in turn represented within theDL formalism to be handled by the DLclassifier. The pivot language has been

especially built for the needs of theapplication, and is not simply anotherXML-based language for documentdescription.

Briefly, the ontology consists of a hier-archy of classes representing concepts,eg genes, and relations between classes,eg interactions between concepts. Eachclass in the ontology is described by a setof attributes and roles representing theproperties of the class.

Two types of classes are available:defined classes with necessary and suffi-cient conditions, and primitive classes,with only necessary conditions.

A document is composed of three parts:(1) a textual abstract; (2) a set of classicmetadata (Dublin core); and (3) a set ofmetadata concerning the content (anno-tations). The pivot language is used forrepresenting the annotations accordingto the ontology, especially the objectsand relations referenced in the docu-ments, in this case, the genes and theinteractions. Objects (instances ofclasses) and relations (instances ofclasses of relations) are described bytheir properties (names and values ofroles and attributes), and the class towhich they belong. The structure of aquery respects the classical schemaSELECT-FROM-WHERE, with someadditional constructs being available.

The classes and relations of the ontologyare translated into concepts within theDL system. Relations are also repre-sented as concepts. All attributes androles are translated into roles in the DLsystem. The two types of classes in theontology — defined and primitive — aretransformed into DL concepts accordingto their status: defined classes becomedefined concepts in the DL system,while primitive classes become primitiveconcepts. However, the properties of arelation, eg reflexivity, antisymmetryand transitivity, must be managed by amodule that is external to the DL system.

Each document is then represented as anindividual within the DL system. Objectsand relations referenced in the ontologyor in the documents are translated intoindividuals within the DL system.Moreover, an individual is related to allthe individuals filling its roles andattributes. Individuals representingobjects and relations are linked to thoserepresenting the documents where theyare referenced.

Classification and subsumption are themain reasoning methods in the DLsystems, and they are used for processingqueries. A query Q can be translated intoone or more query concepts Ci in the DLsystem. Each query concept Ci is thenclassified in the concept hierarchy. Theanswer to the query Q is constituted bythe set of instances of each classifiedquery concept Ci.

by Rim Alhulou and Amedeo Napoli

Representing and handling documents by their content is the core of ESCRIRE,a coordinated action involving three INRIA research teams.

Combining XML and Description Logics for Describing and Querying Documents



The Semantic Web is expected toenhance the development of semanticinteroperability among InformationSystems. The aim is to allowInformation Systems to cooperatewithout requiring modifications to theirsoftware or data organisation. In the ISTEuropean Project ‘Harmonise’, informa-tion interoperability has been addressedwithin a ‘Local As View’ (LAV)approach. The innovative issue of ourLAV solution is represented by the useof a shared ontology to build a commonview of the business sector in which thecooperation takes place.

The Harmonise Approach to Semantic InteroperabilityThe advantages of flexibility and open-ness provided by the Internet in theconnection of computer systems are notmatched in the connection of softwareapplications. The primary techniquesaimed at application interoperability areadapters (typically in EnterpriseApplication Integration) and exchangeformats, such as Electronic Data

Interchange (EDI) or KnowledgeInterchange Format (KIF). Due to thelimited success of existing solutions,Harmonise proposes an approach thatwould start from the above but becentrally based on a domain ontology.To this end, project activities have inves-tigated three main areas.

Interoperability clashes, caused bydifferences in the conceptual schemas oftwo applications attempting to coop-erate. The possible clashes have beenclassified in two main groups:• Lossless clashes, which can be solved

with no loss of information. Examplesinclude naming clashes, when the sameinformation is represented by differentlabels; structural clashes, when infor-mation elements are grouped in adifferent way; and unit clashes, when ascalar value (typically an amount ofmoney, or a distance) is representedwith different units of measure.

• Lossy clashes, which include theclashes for which any conceivabletransformations (in either direction)

will cause a loss of information.Typical cases are information repre-sented at different levels of granu-larity, refinement, or precision. Forexample, in expressing the distance ofa hotel from the airport, one applica-tion simply reports ‘near airport’,while another expresses it in terms ofmiles, eg ‘five miles from airport’.Another example is when one hotelspecifies the presence of an ‘indoorswimming pool’, while another justsays ‘swimming pool’.

Ontology, which represents a common,shareable view of the applicationdomain. This is used to give meaning tothe information structures that are to beexchanged between applications. InHarmonise the ontology is based on theObject, Process, Actor ModellingLanguage (OPAL) representationmethod, which follows the Frame-Slot-Facet paradigm. It includes constructssuch as ISA (with inheritance) andaggregation hierarchies, similarity andvarious kinds of built-in constraints

One of the services that we expect to find with the advent of the Semantic Webis semantic interoperability. The goal of semantic interoperability is to allow the(seamless) cooperation of two Information Systems that were not initiallydeveloped for this purpose. The European Project ‘Harmonise’ aims at buildinga technological infrastructure based on a shared ontology, to enhance thecooperation of European SMEs in the tourism sector.

Harmonise: An Ontology-Based Approach for Semantic Interoperabilityby Michele Missikoff

A number of problems have appearedduring the development of the project.We may underline the following diffi-culties showing the extensions of a KRformalism for representing and handlingdocuments by their content. The DLsystem does not provide any specialconstructor for taking into accountbinary as well as n-ary relations, espe-cially for handling the properties of rela-tions such as reflexivity, symmetry andtransitivity. The possibility of workingwith or without the closed-worldassumption was not available, and wouldhave been very useful. The translations

and the evaluations of queries were notsimple nor always efficient. Actually, amore sophisticated module for handlingqueries based on the formalism ofconjunctive queries has to be practicallydesigned for a realistic document manip-ulation.

The first results of the ESCRIRE projectshow that a DL system such as RACERcan be used with relative success forrepresenting an ontology of a domain,and for describing and querying a set ofdocuments in an XML-like form. Morework must still be done to solve the prob-

lems mentioned above. A number oftheoretical tools exist, but still haveto bemade practical for an effective and real-istic manipulation of documents by theircontents.

Link:http://www.inrialpes.fr/exmo/cooperation/escrire/

Please contact:Rim Alhulou and Amedeo NapoliLORIA, FranceE-mail: {alhulou,napoli}@loria.fr



(such as cardinality constraints, enumer-ation). Based on OPAL, we have devel-oped the ontology management system,SymOntoX.

The concepts in a domain ontology can beseen to be organised according to a(complex) hierarchical structure shapedlike a chestnut (see Figure). In the top part(Upper Domain Ontology) we havegeneric concepts, such as ‘process’,‘actor’, ‘event’ and ‘goal’. In the bottompart (Lower Domain Ontology: LDO) wehave elementary concepts, such as ‘price’,‘streetNumber’, ‘cost’ and ‘internetAddress’. Generally, for two cooperatingpartners, it is relatively easy to reach aconsensus on the concepts of these twoparts. The difficult section is the middlepart — the Application Ontology. Hereconcepts and definitions depend stronglyon the specific application, the kind ofproblems addressed and the method usedto solve them, not to mention the under-lying technology (which often contami-nates the conceptual model) and thecultural aspects. Typical concepts in thislayer are ‘invoice’, ‘customer’, ‘discount’,‘reliableCustomer’, ‘approval’, or moresector-dependent concepts such as‘hotelReception’, ‘confirmedReservation’,‘advancePayment’, ‘lightMeal’ or‘gymTrainer’ (in the tourism sector, forexample).

Semantic annotation, achieved to repre-sent the meaning of a local conceptualschema, expressed by using the conceptsavailable in the ontology. These providea unique semantic reference for eachapplication wishing to expose an inter-face, referred to as a Local ConceptualSchema (LCS) for both exporting andimporting information. In essence, everypiece of information in the LCS will beannotated using the ontology content.For a given application concept (repre-sented as a data structure in the LCS),annotation consists in identifying theLDO elements of the reference ontologyand using them to define its meaning.These semantic annotations are used, atintentional level, to generate themapping between the local conceptualschema and the reference ontology.Mapping rules (among concepts) corre-spond to the transformation rules at

extensional level. Transformations areapplied to local data to code themaccording to the HarmoniseInteroperability Representation (HIR)used to actually exchange data..

ConclusionsThe Harmonise approach to semanticinteroperability is inherently differentfrom the approaches that propose aninterchange format, such as KIF. Theselatter approaches are ‘neutral’ withrespect to the application domain, whileHIR allows information to be exchangedby using ontology terms only. Anotherapproach which is in the line of ourproposal is PIF (Process InterchangeFormat). However, the main differenceis that PIF proposes a format based on apredefined (limited) process ontology,which is given with the standard.Conversely, HIR is based on a richdomain ontology, the content of which isnot provided by the method. It is initiallybuilt by the domain experts, and continu-ously evolves to keep pace with evolvingreality. Accordingly, HIR evolves, sinceits vocabulary is (a subset of) theontology.

Link: http://www.harmonise.org

Please contact:Michele Missikoff, IASI-CNRTel: +39 06 7716 422E-mail: [email protected]

The ontology ‘chestnut’.

UpperDomain Ontology

Application Ontology

Lower Domain Ontology

This project has been running for severalyears at the Institute of Informatics ofSafárik University in Kosice, Slovakia (amember of SRCIM) and involves agroup of institute researchers aided byPhD and Masters students.

The aim of the project was to find waysof increasing the expressivity ofquerying languages by various means:using background knowledge to processboth the queries (formulated in languagethat is as natural as possible) and thesource material, introducing a similarity

measure on the data domains as well asmetadata, and introducing new logicaloperators such as weighted aggregation,which in turn leads to fuzzy or proba-bilistic data models.

by Peter Vojtáš and Jan Vinafi

This report describes a project which was originally applied to information retrievalfrom traditional (deductive) databases, and is now being extended to SemanticWeb sources such as XML, unstructured text, and heterogeneous and distributedsources. The aim of the project is to find effective ways of answering vaguelyformulated queries on crisp, vague, uncertain or incomplete data.

Reducing the Complexity of Information Retrieval by Weighted Querying



While largely successful on the level offormal models and experimentation, thismethod was soon found to create adouble combinatorial explosion. First,there was the data explosion: where adeterministic database consists of a well-defined subset of the Cartesian productof data domains, the fuzzy or proba-bilistic database should contain practi-cally every element of that product(since its probability or truth value willbe positive, though very small).

The second problem deals with theweighting process itself: to assess theprobabilistic measure of a compositeevent consisting of n elementary events,it is in general necessary to evaluate 2nprobability assignments for these events.

It might seem that the first problem atleast could be solved by assigning athreshold value for inclusion in thedatabase. However, it is non-trivial todetermine what partial threshold valueswill lead to a desired threshold for acomposite query (which is necessary ifwe look, eg for k best answers or the so-called e - best answer). This is merelydifficult for fuzzy operators (due to theirspecial nature-truth functionality), butpractically impossible for probabilisticoperators, unless we assume the inde-pendence of the elementary events in thequestion, and certain other restrictions.

Two ways for solving the thresholdproblem were followed in our research.The first way involves the definition andstudy of the fuzzy join of several tablesand different methods of evaluating itwere examined. The phenomenon ofthreshold propagation through the

computational process was studied andvarious heuristics for choosing the orderof join evaluation are now being empiri-cally tested. The second method is basedon a metric induced on the originalattribute domains by a fuzzy linguisticvariable. Consider (see the Figure for avery simple motivating example) thetrapezoidal function ‘sensible price’. Themetric generating function shown in thefigure assigns distance to each pair ofprices. This is easily generalised to adistance between sets. In this way eachrecord (whether with crisp or vaguevalues) is mapped to a set of distanceswith respect to a given elementary query,and from that to a point in n-space,where scales of different axes corre-spond to weights of different attributes(determined by, for example, a neuralnetwork). For choosing the best points(with respect to the ‘ideal’ point locatedin the origin) several methods of compu-tational geometry are available.

Other methods of inducing a metric onthe original data are also possible, suchas that of fuzzy conceptual lattices.

The notion of metric and the corre-sponding similarity measure has otheruses related to IR. One of these uses isthe result of applying the notion to datastructure. Similar data structures aremapped automatically to a unifyingmetastructure allowing the collection ofdata from heterogeneous sources to awhole. This was explored by a groupworking within the project and led to thenotion of the ‘metadata-engine’, whichmay find uses in data mining.

An interesting special case is the miningof data from independently created XMLdocuments. As every author may createhis own ontology to describe themeaning of his XML tags, it is firstnecessary to find a unifying ontology onthe basis of detected similarities.

At present, the loose collaboration ofgroups and individual researchers thatwe have been calling a project is passingfrom the stage of formal models andpromising experimental results to theformation of prototype software tools.This activity is largely being carried outin cooperation with member institutionsof CRCIM, such as the Institute ofComputer Science of the CzechAcademy of Science, the Faculty ofMathematics and Physics of the CharlesUniversity and the Masaryk University.Also participating are our Spanishcolleagues from the University ofMalaga.

The research reported here was partiallysupported by VEGA grant 1/7557/20.

Please contact: Peter Vojtás, P. J. Šafárik University, Košice (SRCIM) Tel: +421 55 62 209 49 E-mail: [email protected]

Metric generation from linguistic variables.



In text, these semantics are often implicitin the text-flow. In multimedia, however,the semantic relations among the mediaitems need to be communicated explic-itly by choosing appropriate layout andstyle. The assumption underlying currentstyle sheet technology (XSLT and CSS),stating that content and presentation areindependent, is often an oversimplifica-tion for multimedia presentations. Inaddition, current style sheets operate onthe syntactic XML level and are unawareof the new generation Semantic Weblanguages, such as RDF(S) and OWL.

Creating well-designed multimediapresentations requires an understandingof both the presentation’s globaldiscourse and interaction structure aswell as the intricate details of multimediagraphic design. In addition, knowledge

of a number of other factors is needed. Adomain description allows the relation-ships among domain concepts to influ-ence the layout and links within thepresentation. Knowledge about theuser’s task or environment allow appro-priate choices of media to be made. Inconjunction with information stored inthe domain model, presentations can begenerated, for example, to skip things theuser already knows and explain newconcepts in terms of already knownconcepts. A description of the character-istics of the end-user platform (such asscreen resolution, bandwidth, ability todisplay colour, audio capabilities) allowsoptimal use to be made of the capabili-ties of the device.

Our prototype generation engine,Cuypers, allows the specification of

these different information types andincorporates them within the overallprocess of generating a presentation. Inorder to make the different knowledgesources explicit for use by the system wewish to use standard languages and tools.Given the range of tools available for theWeb, incorporating Semantic Web toolsand languages was a forgone conclusion.This allows us to re-use, for example,domain descriptions created by expertsin the field, in a language with readilyavailable processing tools. In addition,having generated a presentation, we areable to include semantic mark-up withinit, thus capturing the knowledge usedduring its generation, eg which domainconcepts media items correspond to, orplatform profiles for which the presenta-tion is suited. The current focus of ourwork is on creating a more realistic usermodel, developing a graphic designmodel and investigating the require-ments for a discourse model.

Most work currently being carried outfor the Semantic Web concentrates onthe underlying semantics and is notconcerned with presentation ‘details’.On the other hand, most presentationtools on the Document Web operate onlyon the syntactic level. The long-termgoal is the development of ‘Smart Style’,ensuring that both Semantic andDocument Webs are integrated.

Links:http://www.cwi.nl/ins2/http://www.token2000.nl/

Please contact:Jacco van Ossenbuggen, Lynda Hardman, CWITel: +31 20 592 4141E-mail: [email protected]

by Lynda Hardman

The research goal of the Multimedia and Human-Computer Interaction group atCWI is to investigate the automated generation of Web-based hypermediapresentations tailored to the abilities, preferences and platform of the user. Thisrequires the description and processing of different types of information in orderto assemble semantically annotated media items into a coherent presentation, iea presentation that communicates the intended semantic relations to the user.

Hypermedia Presentation Generation on the Semantic Web

Based on the annotatedmedia database of theRjiksmuseum in Amsterdam,the Cuypers engineautomatically generatesmultimedia presentations inSMIL in response to userqueries. The figure illustratesa presentation about thepainting technique‘chiaroscuro’ in the contextof the work of the painterRembrandt van Rijn. Thepresentation consists of aslide show with examples ofpaintings by Rembrandt

using the chiaroscuro technique, alongside a textual explanation of the technique itself.While consisting of a number of individual media items returned by the database, thepresentation is intended to convey the semantic relations among them.

For example, domain-independent layout knowledge is used to place the presentationtitle centrally, because it applies to both the slideshow and the textual explanation. Incontrast, the title of the textual explanation is left-aligned with the body text to indicatethat it refers only to that paragraph. The close proximity of the text ‘Self Portrait (1661)’to the image, is intended to convey that the text is a label referring to the painting. Inaddition, explicit knowledge of domain-specific presentation conventions allows theengine to display it in such a way that the user can interpret it as being the painting’stitle and the year of creation. This work is carried out in the context of the DutchToKeN2000 project.


In the Semantic Web, content should beadapted to many levels. Some elementsthat influence multi-level adaptabilitycan be identified:• Service providers: parties that produce

Web content, eg yellow pages• Resources: pieces of content, eg

electronic weather reports• Users: parties that consume the

content• Devices: terminals that are used when

accessing the content• Context: situational details to be taken

into account when accessing the content.

Service providers are any parties thatproduce content in the Web, such ascompanies, the public sector, and individ-uals. Resources are individual pieces ofWeb content offered by service providers.Resources can be described with anumber of attributes, such as content type,creator, creation date, price, etc.

Users in turn are any parties that consumethe resources. The most evident users arehuman beings, but some higher levelservices can also be identified as contentconsumers. For example, consider a soft-ware agent collecting weather reportsfrom several information providers in theInternet. This agent consumes the contentand can thereby be functionally identifiedas a user. However, the same agent canact as a service provider to yet higherlevel consumers, such as a Web portalproviding, among other things, informa-tion about weather. As far as human usersare concerned, typical attributes to betaken into account when selecting andadapting content are age, sex, nationality,language, hobbies, etc.

Devices are used by human beings whenaccessing the content. The physicaldetails of devices such as display sizeand amount of memory can have animpact on the adaptation. Consider again

the weather report. When delivered to asmall device with black-and-white char-acter-based display, there is no sense intrying to express the report graphically.At VTT Information Technology, thecontent adaptation based on deviceprofiles has already been studied inearlier projects. Contextual details suchas time and place might also affect theadaptation; for example, a mobile userwould typically prefer the weatherreports of the location he is at or headingto. Kontti-project concentrates particu-larly on the adaptation based on thesecontextual details.

The adaptational relationships betweenthe above-mentioned elements aredepicted in Figure 1. Resources offeredby service providers are adapted to thecharacteristics of different terminals.Contextual details and user preferencesare, however, considered before devicecharacteristics. In this way users receivethe content they wish, in an optimalformat and with situational detailshaving been taken into account.

In the Semantic Web, the most naturalway to express the characteristics of thedifferent elements shown in Figure 1 is inthe form of profiles. Users, serviceproviders, and device manufacturers cancreate profiles in a language, such as RDF(Resource Description Framework),which is recommended by the SemanticWeb community. Depending on theusage, some selected attributes from theprofiles of various elements can bematched in order to deliver personalisedmaterial to the users. Kontti-project aimsto set up a framework where users candesign profiles for themselves anddifferent applications, as well as otherusers being able to utilise these profiles.

The intention in the Semantic Webinitiative is that in addition to human

beings, computer programs can processWeb content, for example to performprofile matching. In order to ensure thismachine-processability and interoper-ability between different independentparties, the profiles should be semanti-cally bound to ontologies. Ontologies,expressed for example in DAML+OIL(DARPA Agent Markup Language +Ontology Interchange Language) or itsforthcoming successor OWL (OntologyWeb Language), specify the meanings ofthe concepts and statements given inprofiles. The weather forecast providermight create a profile for the reportservice using DAML-S — a languagefor describing Web services - on top ofWSDL (Web Services DefinitionLanguage), and store it in a UDDI(Universal Description, Discovery andIntegration) registry for other parties toexploit.

There already exist some standardisationefforts that at least partially cover theelements of the Figure. For example, fordevices there exists CC/PP (CompositeCapabilities/Preference Profiles), forresources Dublin Core, for services theabove-mentioned DAML-S, for usersvCard, for context data Wireless VillagePresence Attributes, etc. While theontologies must be extensible andspecialisable, these standardisationefforts naturally cannot cover all thefeatures people wish to utilise in theirprofiles. In Kontti-project, an ontologyfor expressing the properties required forconstructing contextual profiles iscreated.

Link: http://www.vtt.fi/tte/projects/kontti/

Please contact:Santtu ToivonenVTT Information TechnologyTel: +358 40 839 8405E-mail: [email protected]

One of the key challenges of the Semantic Web is the adaptation of Web contentaccording to diverse parameters. In addition to differences between various devices,contextual information and preferences of users should also be taken into accountin order for a flexible and proactive Semantic Web to emerge. Kontti-project, carriedout at VTT Information Technology, aims to contribute to this vision.

Profile-Based Adaptability in the Semantic Webby Santtu Toivonen

Adaptational relationships between theelements in the Semantic Web.

SPECIAL THEME: SEMATIC WEB



During the electronic publishing processcontent passes through a number ofconversion layers. Each layer modifiesthe content to adjust and customise it inmore detail to create user-oriented valu-able information. The reverse generali-sation process is used to break up docu-ments into basic and original compo-nents such as content, transformationand presentation data. Traditional Webformats (ie HTML) no longer allowreverse engineering of information, as allcomponents are unidentifiably blendedtogether.

The Table shows the four basic layersand exemplarily some of the vocabu-laries available. Each layer correspondsto one distinct stage in the publishingprocess. Vocabulary-specific processingengines are use to handle the data (egXSLT engines to process XML docu-ments with XSLT style sheets).

The physical storage and content data-source is represented by the ContentLayer; the Transformation Layer physi-cally changes the original content byadding, deleting (filter), modifying(process) and converting the content intoa data format supported by the end userapplication. The Presentation Layerdefines and adds style and renderinginformation, used by the application tosense (read, listen, watch) the final publi-cation. The Application Layer finallyrepresents all applications, capable ofpresenting information according tousers’ preferences.

To cater for seamless and automatic inte-gration of local resources into aSemantic Web, content has to be avail-able independently from its transforma-tion and presentation data. XML in

combination with XML-related vocabu-lary specifications (see Table) offers asolid framework to organise and keepcontent, transformation and presentationdata separately and in the most acces-sible way.

The definition of content, transformationand presentation data might be obviousin the context of textual publishing, buthow are these components defined ingraphical, audio and video publications,to become part of a Semantic Web?

GraphicsImages using raster graphics formatssuch as JPEG and GIF are widely used

on the Internet. They are composed of atwo-dimensional grid of pixels, the basicunit of programmable colour. While thismakes information retrieval in terms ofcolour feasible, accessing other imageinformation such as objects, shapes andrelations is ambiguous if not impossible.For purposes of decomposition, imagesusing raster technology are comparablewith badly designed HTML pages,where content and presentation data iswelded together, and it is impossible toextract pure content for furtherprocessing.

One XML-based solution to describingvector graphics is the SVG format, asequence of mathematical statementsthat places lines and shapes into a two-dimensional space. It clearly defines thecontents of the image (mathematicalobjects as well as their grouping) and itstransformation in terms of location, size,applicable filters and deformations. Anoptional CSS file keeps presentation datasuch as colours and margins.

AudioSeparating content from presentationdata in audio files appears to be morecomplex than it is for textual or graphicaldata. Traditionally, audio was distributedto the user as a single source (eg cassette,

CD). Nevertheless, professionalrecording studios use single tracks foreach instrument/voice and compile themtogether before distribution. Oneapproach with an emphasis on audiodelivery over the Internet is MusicXML,which uses XML to describe sheetmusic. Ideally, presentation data such asspeed, volume, key and instruments willbe kept separate from the MusicXMLcontent file. Possible ‘presentationengines’ for MusicXML would include asheet music renderer (for print) as wellas an audio player.

The concept of separating content, transformation and presentation data is oneof the key requirements for constructing a Semantic Web. It has been successfullyapplied in textual publishing for decades (using SGML and DSSSL) and couldfinally revolutionise composition and access of multimedia components such asgraphics, audio and video. What role does XML play in integrating multimediaseamlessly into a Semantic Web?

Integrating Multimedia Components into a Semantic Webby Benjamin Jung and Lyndon JB Nixon

Publication architecture.

Text Graphics Audio VideoApplication Layer XHTML, PDF,

…SVG, CGM,

…MIDI, PDF, … MPEG-4

(BIFS)Presentation Layer CSS, FO CSS Finale templates,

...MPEG-4

(OD)Transformation

LayerXSL, DSSSL, DOM, fxp

Content Layer Various XML vocabularies, (XML-) databases



Making the Web more meaningful andopen to manipulation by software appli-cations is the objective of the SemanticWeb activity. Research at the School ofComputer Applications at Dublin CityUniversity investigates Web compo-nents, which utilise Semantic Web tech-nologies to enable component-basedsoftware engineering (CBSE) for theWeb platform. Efforts in taking the Webfrom a document-oriented to a services-oriented environment — bundled in theWeb Services Framework (WSF) — area first step towards the Web as a softwaredevelopment and deployment platform.The purpose of the WSF is to define adistributed computing model for theWeb and to provide languages and proto-cols for service description, discoveryand invocation. However, addingsemantics to this framework is essential

if Web components instead of Webservices are considered.

Providing reusable software componentsand plug-and-play style software deploy-ment is the central objective of CBSE.Components are reusable, internallycoherent software entities that are speci-fied using interface descriptionlanguages (IDL). Traditional IDLs focuson the syntactical description of servicesthat are provided by the component.However, semantic annotations ofcomponents in form of semantic IDLsare essential for Web components. Inorder to facilitate precise retrieval andmatching and to guarantee the reliableand correct execution of components,semantic service descriptions andcomponent interaction processes need tobe captured in semantic IDLs. Toachieve a high degree of reusability, not

only which services a componentprovides, but also which services acomponent requires needs to be explic-itly specified in an interface. Ourresearch aims to solve problems identi-fied in relation to semantics in the speci-fication of behaviour of componentservices and the component’s interactionprocesses. Here, representing knowledgeabout components is as important asknowledge about the process of compo-nent development.

The Semantic Web objective to open theWeb to manipulation by software appli-cations in our case means to enablecomponent development activities thatare based on the semantical representa-tion of software components. Using theWeb as an infrastructure for componentsdevelopment requires support for twodifferent activities – see Figure 1. Firstly,

Enabling the Web for software components can be ideally supported through theuse of Semantic Web technologies. The Semantic Web activity aims to introducemeaning to the Web using ontologies. Ontologies can create a sharedunderstanding of application domain and development process knowledge thatis crucial for component development activities such as matching and connectingrequestors and providers of component services.

Ontologies for Semantic Web Componentsby Claus Pahl

Video Video is the greatest challenge forcontent and presentation separation.‘Shaped video’ is already used by TVstudios to blend video objects together ;eg in weather forecasts, the weather fore-caster and the background map aregenerated separately. However, objectextraction from the result video stream isdifficult. MPEG-4 is an ISO/IEC stan-dard for the composition of audiovisualscenes consisting of both natural andsynthetic media. While the format isbinary, there is an XML representationfor content interchange. With MPEG-4’sBinary Format for Scenes (BIFS),multiple arbitrarily shaped video objectsmay be composed in an audiovisualscene together with video-qualitysynthetic images. The presentation ofeach video object is defined in an ‘object

descriptor’ (OD) which includesdecoding information, synchronisationand support for scalability. As objects ina BIFS scene can be manipulated indi-vidually, MPEG-4 offers the possibilityof extending media adaptation into therealm of video. Furthermore, MPEG-4has an XML representation for contentinterchange, which allows it to be tightlyintegrated with the Semantic Web.Lyndon’s research is using XML-encoded metadata to dynamicallygenerate MPEG-4 scenes, which, usingshaped video techniques, would makeadaptive video possible.

XLinkXLink, the next generation linking andpart of the XML family of specifications,plays an important role for seamlesslyintegrating multimedia components into

a Semantic Web. Having used the afore-mentioned concepts of separation,XLink allows link sources and targets tobe defined in any granularity, eg rangingfrom a single letter to a paragraph (text),from a single line to the grouping ofobjects (graphics), from a single note to aset of measures (audio) and from a singleframe to an entire scene (video).

Links:http://www.cs.tcd.ie/Benjamin.Jung/http://www.fokus.gmd.de/research/cc/magic/employees/lyndon.nixon/

Please contact:Benjamin Jung, Trinity College DublinTel: +353 608 1321E-mail: [email protected]

Lyndon JB Nixon, Fraunhofer FOKUSTel: +49 30 3463 7103E-mail: [email protected]



discovery and retrieval functionality –possibly in form of a Web componentdirectory or marketplace – allowsproviders of services and components topublish descriptions of their softwareentities and requestors to discover andretrieve them. A notion of contracts tiesmatching components together.However, unlike for a document-oriented Web, a second form of supportis needed. Interaction support allowscomponents to be composed throughconnections for interaction. Essentially,knowledge about all life cycle processesof components is required.

Providing meaning and a shared under-standing are crucial for Web-basedCBSE. Ontologies are means to definethe concepts and properties of a domainand to provide facilities to reason aboutthem. A central objective of our researchis the formulation of ontologies for Webcomponent development. An applicationdomain ontology describes the domainof the software application under devel-opment. A software developmentontology describes the component devel-opment entities and processes. Ontologytechnology integrates notions to describecomponents and their behaviour and aninference system to reason about compo-sition. These ontologies create a sharedunderstanding between software devel-opers and software agents that allow adeveloper to describe provided or

required components and agents toreason about matching of suitablecomponents and to control the composi-tion and interaction of components.

Classical formalisms that we have usedto represent component-related know-ledge are process calculi and modallogics. Ontologies for Web-basedcomponent development, however, haveto adopt Web standards and technolo-gies. DAML+OIL is a Web ontologylanguage. DAML+OIL can be seen as avery expressive description logic.Knowledge representation techniquessuitable for the classical use of the Webare, however, not necessarily directlysuitable for software components andsoftware development, for example if aprocess- or transition-oriented view oncomponents is taken. Our approach toWeb components is based on a descrip-tion logic geared towards the representa-tion of component properties and thesupport of reasoning for matching. Toenable modal reasoning about compo-nents and their behaviour in a descriptionlogic framework is a central objective ofour research. A correspondence betweendescription logics and dynamic logic – amodal logic of programs – that relatesroles and transitions allows us to inte-grate classical software engineeringtechniques into a Semantic Web envi-ronment.

Link:http://www.computing.dcu.ie/~cpahl/CompTech.html

Please contact:Claus Pahl, Dublin City UniversityTel: +353 1 700 5620E-mail: [email protected]

Web Component Life-Cycle. A component request might bymatched by a server componentpublished in a Web componentdirectory. Successful matchingresults in direct communicationbetween client and servercomponent to establish aconnector for service invocationand reply.



The Information Systems on the Webcomply with the semiotic triangle. In theCOEUR-SW triangle, a concept in aUniverse of Thought is related to anuttered symbol in a Universe ofLanguage, the symbol is related to areferent in a Universe of Structure, andthe referent is related back to theconcept. The concept, the symbol andthe referents are related to a context inthe Universe of Discourse (UoD), asembedded in the Web. The COEUR-SWprogram approaches the UoD from threeinterrelated angles and thus seeks tocapture the heart of the Semantic Web.

The Web has three major roles inInformation Systems; one role as thedominant medium for informationdissemination, one as the tool for infor-mation compilation, and one as theevolving information repository. Theroles comprise enterprise and user inter-face issues (UoT), information categori-sation and interpretation issues (UoS),and information storage and informationaccess issues (UoL).

In order to support dissemination andcompilation of information, the reposito-ries may be organised according to theprinciples of integration or interoper-ability. In the former, the relevant infor-mation repositories are viewed as onelarge distributed database, organisedaccording to a common schema. In thelatter, every information system isconsidered to consist of autonomoussubsystems.

The integration principle is dominant inthe classical approaches to informationsystems design. The alternative principleof semantic interoperability is morecompliant with the Semantic Web. Forinteroperable systems to communicatethere must be some agreement on codingas well as meaning of data, but only for

the data that is interchanged. The mutualunderstanding must be as wide as it isnecessary for a particular data inter-change, but need not be wider.

Semantic InteroperabilityData semantics is the relationshipbetween data and what the data standfor. In order to obtain mutual under-standing of interchanged data, the actorshave to share a model of what the datarepresent. Semantic interoperability isabout how to achieve such mutual under-standing.

In order to achieve this we use conceptsand symbols. The concept is the unit ofthought. The symbol is the unit oflanguage. Theories, such as Newton’stheory of motion, are structures ofthoughts, and referents like force, massand acceleration are the units of these

structures. Conceptual knowledge comeswrapped in symbols, eg mathematicalnotation, words or diagrams, which arethe linguistic expressions of knowledge.In order to access the ideas of otherpeople, ie the concepts in their UoT, wemust understand the conceptual struc-tures that are employed in the commonUoS. Furthermore, we need to under-stand the relationship between thesymbols and the ideas they stand for.

Agent SystemsAgents in a multi-agent system are char-acterised by abstraction, interoperability,modularity and dynamism. These quali-ties are particularly useful in that theycan help to promote open systems, whichare typically dynamic, unpredictable,and highly heterogeneous, as is theInternet.

by Arne Sølvberg, Sari Hakkarainen, Terje Brasethvik, Xiaomeng Su,Mihhail Matskin and Darijus Strasunskas

The COEUR-SW program at the Department of Computer and Information Scienceat NTNU in Norway seeks to capture the heart of the Semantic Web.

Concepts on Enriching, Understanding and Retrieving the Semantics on the Web

The COEUR-SW program at IDI, NTNU in Norway, approaches the UoD from threeinterrelated angles. Domain modelling is used in order to capture the knowledge withina UoT into a man/machine understandable theory in a UoS. Ontology Mapping is usedin order to capture the utterances in a UoL into man/machine-retrievable knowledge.Metadata analysis is used in order to capture the theories in a UoS into enrichedman/machine generative utterances in a UoL. In all the approaches inter-relatedconceptualisation of the UoD is used in order to capture the heart of the Semantic Web.



Within a multi-agent system, agentsrepresent their ‘view of the world’ byexplicitly defined ontologies. The inter-operability of such a multi-agent systemis achieved through the reconciliation ofthese views by a commitment tocommon ontologies that permit agents tointeroperate and cooperate. However,between different agent systems, ontolo-gies are developed and maintained inde-pendently of each other. Thus two agentsystems may use different ontologies torepresent their views of the domain. Thisis often referred to as an ontologymismatch. In such a situation, interoper-ability between agents is based on thereconciliation of their heterogeneousviews.

Web Resources Recently, several general-purposemodels for describing Web resourceshave emerged. The intention of theindustry-driven initiatives is to provide ametadata description framework forinterconnected resources. A resource canbe viewed as a bibliographic document(DC, MARC), electronic Web resource(RDF, HTML) multimedia object(SMIL), ontology (OIL, DAML),database concept (MDIS, OIM), or casetool structure (XMI). Many of themodels result from standardisation work

as carried out in industry-driven coali-tions like W3C, MDC, and OMG.

Emerging academic research work hasadapted and applied some of the modelsprovided by the industry coalitions.Metadata is recognised as equally impor-tant for describing the components fromwhich we build services and systems,regardless of whether they are on theInternet or not. The research so far hasfocused on technical aspects of theframeworks, semantic coverage of themodels, and reasoning and queryingmechanisms. The area is becomingmature enough for comparative studiesand prescriptive theories on the proposedmodelling frameworks.

ActivitiesThe overall intention of the COEURSemantic Web research program at IDI,NTNU is to explore and exploit existingresearch results in the areas of concep-tual modelling, information systemsanalysis and design, agent technology,and information systems architectures.Our attacking point is semantic interop-erability. Our main armoury consists ofenriched descriptions and representa-tions for each of the Universes of theSemantic Web triangle. Our ultimatetarget is to enable information systemsdevelopment for Semantic Web applica-

tions in areas such as informationservices, e-commerce, knowledgemanagement, and cooperative systems.

Currently, several activities are underway within this program:

• the Referent Modelling Languageproject, RML

• the Ontology Mapping Using TextCategorisation project, OMUT

• the Semantic Modelling ofDocuments project, SeMDoc

• Conceptual Metadata Analysis andDesign in Information ServiceDevelopment, COMAD

• the Adaptive Distributed InformationService project, ADIS

• the Traceability in CooperativeSystems Development project,CoTracSyDev.

Links:http://www.idi.ntnu.no/coeur-sw/http://www.idi.ntnu.no/coeur-sw/rml/http://www.idi.ntnu.no/coeur-sw/omut/http://www.idi.ntnu.no/coeur-sw/semdoc/http://www.idi.ntnu.no/coeur-sw/comad/http://www.idi.ntnu.no/coeur-sw/adis/http://www.idi.ntnu.no/coeur-sw/cosy/

Please contact: Arne Sølvberg, NTNUTel: +47 7359 3438E-mail: [email protected]

Achieving semantic interoperabilitywhen interconnecting informationsystems of different parties continues tobe a difficult problem. Moreover, a solu-tion that allows for semantic interoper-ability to be achieved at runtime throughdynamic discovery of meaning anddynamic translation between parties isincreasingly needed. In particular, such asolution could have a profound impacton how e-business is conducted. For

example, each party in a B2B relation-ship would conduct business using thenatural terminology of the business,leaving the translation between vocabu-laries to be automatically undertaken bythe system. For example, one partymight call the date for delivery ‘DueDate’ with the format ‘Day/Month/Year’and the other party calls it ‘DeliveryDate’ with format ‘Month/Day/Year’.Such interoperability issues have been

dealt with in the past at schema designtime (with distributed database solu-tions), through the hand-crafting of inter-operability gateways (with system inte-grator solutions) or by forcing each busi-ness to conform to a standard vocabulary(eg ebXML). Of course, the difficultycomes as the nature of each business andtheir respective ontologies evolve overtime. More and more what is required isthe runtime comparison of ontologies

In Multimedia and Hypermedia Systems the Knowledge and Data EngineeringGroup at Trinity College Dublin has pioneered research in virtual Web-basedenvironments, adaptive hypermedia systems for customised or personalised userexperience, usability of Web-based learning, metadata-driven (XML based)approaches for developing re-usable digital resources and XML-based digitalresource discovery services in a global search context.

Negotiating on Your Own Terms!by Declan O’Sullivan and Vincent Wade



and the dynamic construction of gate-ways.

Research that is being undertaken withinthe Knowledge and Data EngineeringGroup of Trinity College Dublin looks atfinding such a solution, and draws uponongoing research into the Semantic Web,ontologies, electronic negotiation, anddynamic adaptation of software systems.The chosen application area, whichprovides a rich source of requirements, isthe management of ubiquitouscomputing/smart spaces environments.A Smart Space is a physical space rich indevices and software services that iscapable of interacting with people, thephysical environment and externalnetworked services. The aim of theSmart Space is to bring tangible benefitsto people in support of their tasks andactivities. Take for example, a cardiacsurgeon who is woken by her alarmclock. Upon sensing that she has got upand is having a shower, the home smartspace sends messages to the coffeemachine, toaster and fruit press appli-ances to have breakfast ready. Her PDAis activated and downloads any emer-gency case patient files that may havearrived overnight, and alters the day’sappointment and work schedule ifnecessary. Upon leaving the house, thehome management system configuresthe intruder alarm, the callforwarding/answering service and utilitymanagement services accordingly. Theprimary objective of smart spacemanagement is the dynamic runtimeadaptation of smart space devices andsoftware services to provide thenecessary support for people’s tasks andactivities. In addition it involves the life-cycle and operational management of thedevices and services offered by the smartspace.

Today’s computers interact with users interms of low level abstractions – applica-tions and individual appliances. A smartspace will take over most low levelmanagement of applications and appli-ances so that the user interacts in termsof ‘activities’ (that represent user intent).In this scenario, it cannot be assumedthat ordinary users will limit theirexpressions to a set of standardisedactivities. Indeed this would run contraryto the intent of smart spaces to support

users to perform their activities seam-lessly and efficiently. Where a user isconfined to a single operator domain, itcan be foreseen how the smart space canlearn which user-specified activities canbe mapped to which smart spaceservices, and which of those services canbe executed simultaneously withoutconflict at runtime, in accordance withpolicies set for that space. However, inorder to enable the free movement ofusers between smart spaces, informationmust be made available to the visitedsmart space regarding which equivalentor adapted services the user will want toavail themselves of, and what quality ofservice will be expected/agreed to. Eachsmart space will want to negotiate usingthe activity/service and contractingterminology that is most familiar. Thisresults in the need to ‘bridge’ betweenthe terminology of each party during thenegotiation of the necessary service levelagreements. However, given the widediversity of users, smart space capabili-ties and smart space operators envisagedin the future, the negotiation of suchservice level agreements needs to beautomatic, with human interventionbeing a very exceptional case.

One solution being researched in thegroup involves a platform that usesSemantic Web technologies to dynami-cally bridge the terminologies of the twoparties at runtime. First the ontologies of

the two parties (represented in XMLTopic Maps) are compared. The compar-ison of the ontologies is then used toderive XML stylesheets that transformthe negotiation from one party’s termi-nology to the other party’s terminologyand vice versa. This comparison anddynamic bridging is undertaken atruntime and requires no a priori know-ledge of the party requesting the servicelevel agreement. Key to our approach isthat each ontology needs only to adhereto a small subset of common concepts,with the rest of the ontology beingdefined to reflect the terminology mostnatural for the particular smart space.The approach is illustrated in the Figure.

This work is undertaken as part of theIrish government HEA funded M-Zonesproject. The M-Zones ResearchProgramme is a multi-disciplinary, inter-institutional research programme thatengages in fundamental research inManagement and Control systems forintegrating Multiple Smart Spaces.

Link:http://kdeg.cs.tcd.ie

Please contact:Declan O’Sullivan or Vincent Wade Trinity College DublinTel: +353 1 608 1765E-mail: [email protected], [email protected]

Overview of Smart Space Mobility Negotiation tool.

Visitor Smart Space A

TargetSmart Space B

OntologyA

Requests interms of A

Response interms of A

OntologyB

Ontologycomparision

TransformerGenerator

Requests interms of B

Response interms of B

Analysis

creates

RuntimeTransformer



Articles in this Section45 Visualisation Middleware for E-Science:

A New Project in the UK E-Science Programmeby Julian Gallop, CLRC

46 Dealing with Massive Volumetric Visualisation:Progressive Algorithms and Data Structuresby Rita Borgo, Pisa University and CNR,Valerio Pascucci, Lawrence Livermore National Lab, USA,and Roberto Scopigno, ISTI-CNR

48 RADIUS: Looking for Robots’ Helpin Computer Science Research and Educationby Jozef Kelemen and Aleš Kubík, Silesian University, Opava, Czech Republic

49 Consensus Creation on Universal Access in Health Telematicsby Constantine Stephanidisand Demosthenes Akoumianakis, ICS-FORTH

50 A Body-Monitoring Systemwith EEG and EOG Sensorsby Mária Bieliková, Slovak University of Technology (SRCIM)

52 WebRemUSINE: A Tool for Usability Evaluationof Web Applicationsby Laila Paganelli and Fabio Paternò, ISTI- CNR

53 Digital Factories, Production Networksby László Monostori, Géza Haidegger, József Vánczaand Zsolt János Viharos, SZTAKI

54 GenoStar: A Bioinformatics Platform for Exploratory Genomicsby François Rechenmann, INRIA



Computer-assisted visualisation isalready a key technology for under-standing large-scale simulations andobservations and, when appropriatelyconfigured, will enable collaborations ofresearchers to analyse the results ofevolving e-scientific resources. To besuccessful at this, visualisation tech-nology needs to be adapted to theevolving Grid, which underpins e-science. gViz, funded by the UK E-science Core Programme started onAugust 2002, aims to research anddevelop visualisation middleware for e-science.

SituationsUsing the Grid presents new situationsfor anyone developing visualisation soft-ware:• An intensive computation on the Grid

may generate large results datasets. Tobe viable, some visualisation algo-rithms would also need to be run onthe Grid.

• Visualizing a large distributed datasetor multiple heterogeneous ones

requires access to Grid data manage-ment interfaces.

• The existing models for collaborationneed to be extended. A virtual organi-sation may involve dynamic collec-tions of people who may alternatebetween virtual meetings and asyn-chronous communication. Also thevirtual organisation may not be able toenforce one specific visualisationsystem.

• There will be a need to integrate visu-alisation software with emerging Gridtechnologies, such as GSI to handleauthentication, Grid replica manage-ment to handle large data files andother Globus facilities.

• Visualization frameworks — whetherdataflow networks such as IrisExplorer or interpretative program-ming languages such as PV-Wave orJava programming systems such asVisAD — can include computationalmodules when the whole framework ison a single processor or a fixed,limited distributed system. This leadsus to ask whether there is a possibility

that the framework can control compu-tation modules executing on the Grid.

Work PackagesThere are three main subdivisions ofwork:• two workpackages aim to grid-enable

two very different visualisationsystems, Iris Explorer and pV3, whichare both exemplars of useful systemclasses

• another workpackage aims to developcompression techniques suitable fortransmitting very large streams of data

• the remaining workpackage aims tobuild up appropriate XML languagesfor describing the structure and formatof data to be analysed and forprocessing tasks. In addition to thetechnical definitions, this workpackagewill, early in the project, run a work-shop on XML for visualisation.

The Figure shows some earlier workcarried out at the University of Leeds(the lead partner in the new project) aspart of a previous UK e-science demon-strator project (see http://www.visualiza-tion.leeds.ac.uk/CovisaG/). This showsat a basic level that it is possible tocommunicate with the Grid using theframework provided within an off theshelf visualisation system. The top of thefigure shows some modules connectedtogether, two of which are wrappers forGrid services. The bottom of the figureshows a rendering window which showsa case study of hazardous pollutants.This work was demonstrated at theopening of the UK National e-ScienceCentre in April. The new project willneed to go on to exploit Grid capabilitiesincluding security, flexibility, symmetrybetween collaborators and extensibility.

E-science is intended to enable global collaborations over the Internet and provideaccess to very large, distributed resources, including data collections, computationand networks of specialised experimental facilities. A new project, funded by theUK E-science Core Programme aims to research and develop visualisationmiddleware for e-science.

gViz: Visualisation Middleware for E-Science:A New Project in the UK E-Science Programmeby Julian Gallop

Screenshot of a session using the Iris Explorer visualization system accessing a Grid application.



Also relevant to the new project isprevious work on component-orientedvisualisation carried out by CLRC RALand the University of Lancaster (projectVisual Beans (http://www.acu.rl.ac.uk/VisualBeans/).

ApplicationsThe intention is to develop techniquesand tools which are usable across a widerange of applications. However withinthe project, certain applications will beused as testbeds and these include heartmodelling (in computational biology)and climate prediction.

PartnersThe academic partners are the Universityof Leeds (lead partner), the University ofOxford, Oxford Brookes University andCLRC Rutherford Appleton Laboratory.

In accord with the funding rules, theindustrial partners make, in kind ordirectly, a total contribution whichmatches the funding received from the e-science programme. These partners areIBM UK, Nag, and Streamline Computing.

FollowupThis new project may be of interest tothose who are developing Grid applica-

tions which require visualisation or thosewho are developing more advanced Gridtools and believe that visualisation soft-ware may make use of them. It would beuseful to hear from European researcherswho are interested in this project.

Links:http://www.bitd.clrc.ac.uk/ and follow links to R&D then Visualization Middleware.

Please contact: Julian Gallop, CLRCTel: +44 1235 44 5569E-mail: [email protected]

Projects dealing with massive amounts ofdata need to carefully consider all aspectsof data acquisition, storage, retrieval andnavigation. The recent growth in size oflarge simulation datasets still surpassesthe combined advances in hardwareinfrastructure and processing algorithmsfor scientific visualisation. As a conse-quence interactive visualisation of resultsis going to become increasingly difficult,especially as a daily routine from adesktop. The visualisation stage of themodelling-simulation-analysis activity,still the ideal effective way for scientiststo gain qualitative understanding of theresults of simulations, then becomes abottleneck in the entire process. Thisproblem poses fundamentally new chal-lenges, both to the development of visu-alisation algorithms and to the design ofvisualisation systems.There is a need atsystem level to design the visualisationprocess as a pipeline of modulesprocessing the data in stages. In this waya flow of data is created that needs to beoptimised globally with respect to magni-tude and location of available resources.To address these issues we have been

working on a new progressive visualisa-tion algorithm that subdivides the inputgrid following regular criteria and thentraverses the regular structure producedextracting the cells containing isovalues.These cells are then organised in a hierar-chical structure (from coarse to finelevels) and subsequent levels of detail areconstructed and displayed to improve theoutput image.

We separate the data extraction from itsdisplay. The hierarchy is built by asingle process that traverses the input3D mesh. A second process performsthe traversal and display. The schemaallows us to render partial results at anygiven time while the computation of thecomplete hierarchy is progressing. Theregularity of the hierarchy makes itpossible to create a good data-parti-tioning schema that allows us to balanceprocessing time and data migrationtime. The subdivision criteria is itselfsuitable for parallelisation. The algo-rithm works in the following way:

Subdivision: we generate a partition ofthe input data. Figure 1b shows the 2Dedge-bisection refinement sequence thatwe adopt to sub-sample a 2D regulargrid. The coarse level is basically trian-gular mesh. Each refinement step insertsa new vertex on an edge and splits theadjacent triangles along this edge intotwo. Instead of ‘triangles’ we reason interms of ‘2D diamonds’ consideringeach cell as the composition of the twotriangular halves that will be split by thebisection. For the 2D case we can subdi-vide the diamonds into two main classes:first class square-shaped diamonds (ortype 0 diamonds, Figure 2a), and secondclass rhombus-shaped diamonds (or type1 diamonds, Figure 2b). Each diamond ischaracterised by a centre (the centre ofthe bisection edge).

In 3D, the bisection edge refinementbecomes a schema for the subdivision oftetrahedral meshes. It applies to tetrahe-dral meshes in the same way that itapplies to 2D triangular meshes. Withrespect to the 2D case it maintains all theproperties. Each cell is still subdivided

This article summarises recent results obtained in the area of visualization ofmassive datasets exploiting the strength of subdivision techniques and distributionalgorithms. The focus is on trying to meet storage requirements designingvisualisation algorithms suitable for distributed environments.

Dealing with Massive Volumetric Visualisation:Progressive Algorithms and Data Structuresby Rita Borgo, Valerio Pascucci and Roberto Scopigno



along ‘the longest edge’, which corre-sponds to the main diagonal of the cellitself. Each bisection gives birth to anarbitrary number of diamond-like newcells. For regular grids, the starting cell isa first class cube-shaped diamond cell(see Figure 3a). The bisection of a firstclass diamond cell gives birth to secondclass octahedral-shaped diamond cells(Figure 3b-c). Each new second classdiamond cell (Figure 3d) is bisectedalong its main diagonal andthird class esahedral-shapeddiamond cells are created (seeFigure 3e-f). Each esahedralcell (Figure 3g) is bisectedalong its longest diagonal andgenerates first class diamondcells (Figure 3h-i). In 3D, weneed to repeat the bisectionprocedure three times to reacha stage in which a new bisec-tion gives birth to first classcube-shaped cells, returning tothe subdivision step for thestarting cell (see the third leveldiamond in Figure 2e) .

Hierarchical Organisation:The subdivision schemaimplies a hierarchical organi-sation of the dataset. It is easyto organise such a hierarchy ina tree-like data structure andto extract a sort of ‘seed set’,made up of cells whoseinternal range of isovaluesincludes the isovalue target.The seed set generated corre-sponds to an adaptive traversal

of the mesh at different levels. The hier-archy always starts with a first classdiamond cell and proceeds through eachlevel of refinement with an alternation ofsecond, third (for the 3D case) and firstclass diamond cells.

Isosurface Extraction: Each generateddiamond internally holds a ‘piece’ of thetotal isocontour that needs to beextracted. To perform the extraction, for

the 3D case, we subdivide each diamondcell into tetrahedra and apply a marchingtetrahedra algorithm. Each isocontour isupdated within a single tetrahedron andthen composed to update the globalisosurface within the set T of all tetra-hedra around the bisection edge.

The key novelty of the present schema isthat, by providing a set of local rules forcontinuous geometric transitions

(geomorphs) of one level ofresolution into the next, webridge the gap between adap-tive techniques and multi-resolution decimation-basedtechniques. Moreover, theregularity of the schemapermits the design of an effi-cient run-time data parti-tioning and distribution algo-rithm which will reduce localmemory requirements and willmake use of the potential of thecurrently under-exploiteddistributed environment.

Please contact:Rita BorgoPisa University and ISTI-CNRTel: +39 050 315 3471E-mail: [email protected]

Roberto Scopigno, ISTI-CNRE-mail: [email protected]

Valerio PascucciLawrence Livermore National Lab, USAE-mail: [email protected]

Figure 1: a) 2D First and second level diamonds; b) 2D subdivision. Figure 2: 2D diamonds vs. 3D diamonds.

Figure 3: First, second and third level diamonds and respectivesubdivision and fusion with adjacent cells.



RADIUS, based on the experience of theproject participants in such varyingscientific areas as formal languagetheory, artificial intelligence, artificiallife, natural computing, multi-agentsystems and agent-based economies,reflects this paradigmatic change in twoways; research in robotics and decen-tralised intelligence underlying systems,and the incorporation of ideas fromstudies in this field into the computerscience and applications curricula.

Our results and experiences in the theoryof grammar systems (GSs), a recent areaof formal language theory, have proventhat formal languages can also bedescribed and studied in terms of (finite)collections of grammars rather than indi-vidual grammars. This theory providesus with an effective framework forresearch in fields such as multi-processing, decentralisation, cooperationand emergence of collective phenomena,and can thus be viewed through the lensof the above new computationalparadigm.

In addition to research, grammar systemsas a part of Masters and doctoralcurricula have been presented by Alica

Kelemenová and Jozef Kelemen atCzech and other European universities(Silesian University, Opava, Universityat Hradec Králové, both Czech Republic,University of Economics at Bratislava,Slovakia, Vienna University ofTechnology, Austria, and Rovira iVirgili University, Tarragona, Spain). Atutorial was delivered by Jozef Kelemenon the connection of a new, non-func-tional modularisation of systems andgrammar systems at the 9th ECCAIAdvanced Course on AI in Prague, 2-13July 2001.

We have supervised Masters and PhDtheses in GSs; for example, a formalframework and computer simulation as atool for experimentation have been usedin order to prove emergence of somephenomena in agent-based computa-tional economies in the PhD dissertationof the second author of this article.

Besides investigating multi-agentsystems as grammar systems, we haveexperience with teaching agent-basedsoftware engineering and the theory andpractice of multi-agent systems as aregular Masters-level university course.A collection of lecture notes prepared byAles Kubík and published by the SilesianUniversity for this purpose was the firstcomprehensive introductory textbook onthe topic published in the Czechlanguage.

In spring 2002, we set up a laboratory forexperimentation in information-processing aspects of collective robotics.It is equipped with K-team Kheperarobots. We study societies of robots asopen information systems of physicallyembodied agents executing tasks in ashared dynamic environment. For

programming the robots we use C andJava, with a Lisp interface for control-ling them remotely.

RADIUS is focused on studying theemergence of complex behaviours fromthe interactions of simple agents, inorder to make emergence more tractablethrough the design and implementationof multi-agent architectures on the theo-retical ground provided by grammarsystems. Proven capabilities of formaltools can thus be experimentally tested inreal environments. In conjunction, westudy how the behaviour of a group ofrobots can evolve from an initial specifi-cation, and how to design robotcontrollers using neural nets, interactiveTuring machines, biologically inspiredcomputing etc.

Their modularity and easy-to-use inter-face make Kheperas suitable forteaching distributed programming andthe information-processing base ofrobotics, and for teaching basiccomputer programming in a non-tradi-tional manner, stressing the ideas ofStein’s new modularity rather than thetraditional functional modular decompo-

by Jozef Kelemen and Aleš Kubík

RADIUS is a project at the Institute of Computer Science at the Silesian Universityin Opava, Czech Republic, which reflects the shift from viewing computation asan execution of a sequence of basic computational steps, towards viewing it asthe activity of a community of interacting basic computational entities. Theacronym RADIUS is inspired by the robot of the same name from the play ‘R.U.R’(Rossum’s Universal Robots) by the Czech writer Karel Capek.

RADIUS: Looking for Robots’ Help in Computer Science Research and Education

Figure 1: The Khepera robots.Figure 2: The authors playing with the robots.



sition. By programming robots, studentsmust interact with autonomous units thatare driven by the controller, with severalprocesses with ‘endless’ loops thatcompete for the robot control. Studentsface the problem of programming realdevices equipped with sensors and effec-tors which interact with the dynamicenvironment. This helps them to realisenot only the particular characteristics ofprogramming open information systems,but the limitations – which from ourperspective is the most important point –of the object-oriented analysis anddesign of information systems incomparison with agent-based analysisand design of distributed applications oreven robotic societies.

It is intended that RADIUS will continueto be used in the formulation and solu-tion of further problems of new modu-larity, collective robotics, artificial life,computational economies and otherfields within the area of grammarsystems. It will also find employment inthe experimental verification ofemerging hypotheses (we plan to buildour own peripherals in order to enhancethe sensory/effectory capabilities ofrobots), and in incorporating new ideasinto our programming courses, makingthem more effective, attractive and inter-esting for our students.

Looking forward to further commonwork in order to achieve these goals, we

acknowledge the long-term, intensiveand fruitful cooperation with our part-ners, especially with the Research Groupon Modelling Multi-Agent Systems,SZTAKI, and the above-mentioneduniversities.

Link: http://ui.fpf.slu.cz

Please contact:Jozef KelemenSilesian University, Opava, Czech RepublicTel: +420 653 684 361E-mail: [email protected]

Ales KubikSilesian University, Opava, Czech RepublicTel: +420 653 684 367E-mail: [email protected]

The objectives and technical approach ofthe IS4ALL project have been presentedin ERCIM News Nos. 46 and 43. Thisarticle focuses and elaborates on theproject’s activities that aim to facilitatetransfer of know-how and know-whyregarding Universal Access to potentialrecipients in the Health Telematicsindustry.

To this end, the perspective of theIS4ALL project is three-fold. Firstly,there is a commitment to consolidate andcodify currently available knowledgeand best practice on Universal Access bymeans of a continuous data collectioneffort. In a second stage, methods andtechniques for Universal Access are vali-dated by reference to designatedscenarios and use cases from the domainof Health Telematics. Finally, a broadrange of outreach activities are plannedto facilitate awareness raising andconsensus creation on the issue of

Universal Access amongst HealthTelematics practitioners. The primetarget of these outreach activities is tofacilitate inter- and intra-sectoral transferof knowledge with the aim to: • increase Health Telematics practi-

tioners’ propensity to innovate as aresult of process improvementsresulting from the adoption ofUniversal Access practices

• guide them towards the appropriationof the benefits of innovative applica-tions and services following theprinciples of Universal Access.

In devising suitable mechanisms to facil-itate transfer of know-how and know-why, the commitment is towards supportmeasures intended to promote in theHealth Telematics sector a favourableenvironment for the exchange of know-ledge, experience and practice on theissues pertaining to Universal Access.The prime focus is on mechanisms that

foster networking and learning by doing.Two such clusters of mechanisms havebeen defined to facilitate collaborationand dissemination of information,respectively.

Collaboration-Oriented MechanismsIS4ALL promotes active collaborationwith players in the Health Telematicssector by: • Formally inviting key representatives

from the Health Telematics sector tobecome involved as partners in theIS4ALL consortium. In this way, theIS4ALL project has acquired stronglinks with user communities, such asMS-HUGe, as well as industrialnetworks such as EHTEL. Addition-ally, these representatives provide theproject with the collective wisdom,requirements and needs of the HealthTelematics sector, thus focusing tech-nical work and ‘screening’ the results

by Constantine Stephanidis and Demosthenes Akoumianakis

IS4ALL (IST-1999-14101) ‘Information Society for All’ is an EC-funded ThematicNetwork seeking to establish a wide, interdisciplinary and closely collaboratingnetwork of experts (Working Group) to provide the European Health Telematicsindustry with a code of practice detailing how to appropriate the benefits ofUniversal Design.

Consensus Creation on Universal Access in Health Telematics



The Body-Monitoring System (BMS) isdesigned as a mobile device that is ableto collect measured data and to actaccording to instructions set by a super-visor. The system consists of a body-monitoring network (see Figure 1). Inorder to recognise the monitoredperson’s state, the monitor unit connectsto various body sensors and i/o devicesusing either wired or wireless communi-cation technologies. Data from allsensors is collected, stored and analysedin real time and, according to the anal-ysis, actions may then be performed. Acomputer is used as an interface to thebody-monitoring network, and devel-

oped software allows a supervisor toconfigure the monitor unit for the moni-tored person, to connect sensors and i/odevices, define and upload instructionsfor monitoring and download collecteddata.

The monitor unit software consists of acommunication module (responsible forconnecting and controlling sensors, andfor gathering and pre-processingmeasured data), a storage module (forstorage of collected data), and a policyinterpretation module (responsible forcontrolling the behaviour of the monitor

unit according to instructions defined bya supervisor).

Two types of drivers are introduced. Therole of a communication driver is to hidethe way in which data is transmitted.There is one driver for every type ofcommunication interface, eg a Bluetoothdriver or an IEEE 802.11b driver. Thecommunication driver does not careabout the data itself; this is the role ofdevice drivers. Each type of sensor hasits own device driver. When a devicedriver receives a message from one of itssensors it decodes the message andinforms the policy engine about the state

A team of four students from the Slovak University of Technology in Bratislavahas developed a prototype mobile sleep laboratory — a body-monitoring systemwith EEG and EOG sensors. The students explored the concept of intelligent datacollection from human body sensors, and with this system, won third prize at theannual IEEE Computer Society International Design Competition (CSIDC 2002)World Finals.

A Body-Monitoring Systemwith EEG and EOG Sensorsby Mária Bieliková

so as to ensure their relevance,comprehensiveness and applicability.

• Initiating and sustaining directcontacts with target user communities,which may be national HealthTelematics providers and networks,industries, or consortia, in an effort toelicit requirements and formulatereference scenarios, so that UniversalAccess methods and techniques arevalidated. At the same time, the targetuser communities gain feedback ontheir technical developments in theform of usability reports, recommen-dations for design improvements,participation in seminars and access toinformation and project deliverables.

• Organising thematic meetings in theform of Focus Groups or workshops.Experts are invited to review progressand technical achievements, and tohelp the IS4ALL project consolidateand refine its findings and fine-tunethe presentation of its technicalachievements. Thus far, the project hasorganised one Focus Group seeking tocollect information on Universal

Access methods, techniques and tools,while three workshops are planned totake place by September 2003 to refineand finalise the project’s final deliver-able, which is a comprehensive codeof Universal Access practice in HealthTelematics.

Information Dissemination MechanismsThe outreach activities of the IS4ALLproject amount to organising andextending invitations for participation toseminars, tutorials and workshops,preparing and periodically updatingproject presentation materials, and main-taining the project’s website. Seminarsconstitute a prime instrument for dissem-inating information in a timely fashionand are delivered, free of charge, in thecontext of important events organised bymembers of the Health Telematics sector(ie European and international confer-ences, workshops, etc).

IS4ALL has already delivered threeseminars, the content of which varied to

reflect progress and technical achieve-ments, focusing respectively on:• presenting Universal Access methods

as adopted by practitioners in othersectors of the industry, such as indus-trial engineers, architects and devel-opers of interactive applications andservices

• elaborating Health Telematics refer-ence scenarios to ensure their validityand relevance

• illustrating the methods in practicethrough designated validation exercises.

IS4ALL plans to organise three moreseminars by September 2003. For moreinformation on the IS4ALL project andscheduled activities, please visit theproject’s website.

Link:http://is4all.ics.forth.gr

Please contact:Constantine Stephanidis, ICS-FORTHTel: +30 810 391741E-mail: [email protected]



of the sensor. To send/receive a messageto/from a sensor, the device driver usesthe corresponding communicationdriver.

The behaviour of the monitor unit iscontrolled by interpretation instructionsdefined by a supervisor (called policies).A policy describes the monitor unit’sresponse to events reported by sensors.Policies are written in the Policy MarkupLanguage (based on XML). It enablesgeneral policies to be written using‘virtual’ objects that are replaced byphysical devices in actual runtime. Themonitor unit prototype is implemented inMicrosoft-embedded Visual C++ 3.0 onCompaq iPaq Pocket PC HR3870.

To validate the system design thestudents tested it in a specific field ofmedicine – sleep research. To cope withthe problem of sleep disorders, sleeplaboratories in hospitals are used tomonitor patients overnight. However,patients are influenced by the hospitalenvironment, and usually show differentsleep patterns to patients at home. As asolution to this problem a prototype

mobile sleep laboratory was developedfor home use. The prototype employs anelectroencephalograph (EEG, whichmonitors brain waves), an electrooculo-graph (EOG, which monitors eye move-ment) and a thermometer. Analysis ofEEG and EOG data allows identificationof all sleep stages.

Sensor implementation goes out from acommon sensor platform designed in thecourse of this project. The common plat-form contains a detecting element,amplifiers and filters, an AD converter, amicroprocessor and a Bluetooth module.Figure 2 depicts the EEG sensor proto-type. The developed EEG sensor couldalso be used for continuous EEG exami-nation. Such an examination is necessaryfor patients suffering from epilepsy.

The most distinctive features of thedescribed project are: • it provides wireless communication

between various sensors attached tothe human body

• it provides a unified way of controllingseveral sensors by a single monitordevice. Sensors may also influenceeach other, eg it is possible to startmonitoring blood pressure using amanometer as soon as an ECG (elec-trocardiograph) sensor reports a heartproblem

• it is possible to adapt monitoring to thepatient’s state, ie measurements andalert messages are controlled andprovided according to the currentcontext

• it allows complex schemas for sensorcontrol to be defined and representedin a specialised language called PML(Policy Markup Language), which ispowerful enough to describe mutual

dependencies and cooperation amongseveral sensors, and even to definecomplex schemas for taking medicine

• it provides a secure way to transmitand store measured data

• it can be used for monitoring almostany interesting body parameters, sincethe design supports different sensorsusing practically any communicationtechnology.

Although the students concentrated onmonitoring the human body for medicalpurposes, the design of the Body-Monitoring System could also be used inmany other fields (eg pulse rate moni-toring in sports science, prevention ofSudden Infant Death Syndrome andmonitoring of people working indangerous environments).

AcknowledgementI would like to thank our students PeterBlstak, Matus Horvath, Peter Lacko andMarian Lekavy for their excellentproject work and presentation at theCSIDC World Finals in Washington,DC. The work reported here waspartially supported by the SlovakScience Grant Agency, grant No.G1/7611/20, and by the ComputerSociety IEEE via the Computer SocietyDesign International Competition 2002.

Links: SUT CSIDC web page:http://www.dcs.elf.stuba.sk/csidc

CS-IEEE CSIDC web page:http://www.computer.org/csidc

Please contact:Mária Bieliková Slovak University of Technology (SRCIM)Tel: +421 2 602 91 473E-mail: [email protected]

Figure 1: Overview of the Body Monitoring System.

monitor unit

i/o device body monitoring network

sensor

Figure 2: EEG sensor protoype.

Figure 3: The EOG sensor in action.



We have developed a method and anassociated tool to detect usability prob-lems in Web interfaces through a remoteevaluation where users and evaluatorscan be separated in time and space. Ourapproach combines two techniques thatare usually applied separately: empiricaltesting and model-based evaluation. Thereason for this integration is that modelscan be useful to detect usability prob-lems but their use can be much moreeffective if they can be related to theactual use of a system. Our tool is able toanalyse the possible inconsistencybetween the actual user interactions andthe task model of the Web site thatdescribes how its concrete designassumes that activities should beperformed. To support remote evalua-tion, we have developed a technique thatmakes it possible to record user actionsduring a site visit. The analysis of the logdata is based on a comparison of thetraces of actions performed with thestructure of the task model. This analysisprovides evaluators with a number ofresults that are related to the tasks that

users intend to perform and the Webpages and their mutual relationships.

In our case we follow a hybrid approachbecause our environment is able toanalyse data relative to user interactionsand then compare them to the task modelcorresponding to the design of the Website. The solution that we have adoptedin WebRemUSINE to identify userintentions is to display the high-leveltasks that are supported by the Web siteasking the user to indicate explicitlywhat task they want to perform.WebRemUSINE compares the logs withthe task model and provides resultsregarding both the tasks and the Webpages supporting an analysis from bothviewpoints.

The method is composed of three phases:preparation, which consists of creatingthe task model of the Web site, collectingthe logged data and defining the associa-tion between logged actions and basictasks; automatic analysis, whereWebRemUSINE examines the logged

data with the support of the task modeland provides a number of resultsconcerning the performed tasks, errors,loading time; and evaluation, the infor-mation generated is analysed by the eval-uators to identify usability problems andpossible improvements in the interfacedesign.

The environment is mainly composed ofthree modules: the ConcurTaskTreeseditor CTTE) (publicly available athttp://giove.cnuce.cnr.it/ctte.html); thelogging tool that has been implementedby a combination of Javascript and Javaapplet to record user interactions; andWebRemUSINE, a Java tool able toperform an analysis of the files generatedby the logging tool using the task modelcreated with CTTE.

Link:http://giove.cnuce.cnr.it/webremusine.html

Please contact:Fabio Paternò, ISTI-CNRTel: +39 050 315 3066E-mail: [email protected]

by Laila Paganelli and Fabio Paternò

Creating a Web site allows millions of potential users with various goals andknowledge levels to access the information that it contains. For this reason,interest in usability evaluation of Web sites is rapidly increasing.

WebRemUSINE: A Tool for Usability Evaluation of Web Applications

WebRemUSINE interface.



Manufacturing systems of our epochwork in a rapidly changing environmentfull of uncertainties. Besides internalfactors such as malfunctions and break-downs, the main external reasons foruncertainty are:• fast-increasing and diversified

customer demands• the growing role of one-of-a-kind

production, and rapid sequences ofnew tasks

• an increase in the number and speedof communication channels

• the appearance of new technologies• changes made by partners (suppliers,

distributors, customers, purchasers)• market instability (eg the sudden

changes of raw material prices).Increasing complexity is another charac-teristic that is evident both in productionprocesses and systems and in enterprisestructures.

The concept of the digital enterprise, thatis, the mapping of the importantelements of enterprise processes bymeans of information technology tools,gives a unique way of managing theabove problems. However, the optimalor nearly optimal exploitation of thehuge amount of information available isimpossible without the effective applica-tion of the methods and tools of artificialintelligence. In this way, systems can bedeveloped that are expected to solve,within certain limits, unprecedented andunforeseen problems on the basis of evenincomplete and imprecise information.This is the early concept of IntelligentManufacturing Systems (IMS) by thelate József Hatvany, from 1983.

The project partners for ‘DigitalEnterprises, Production Networks’,which is supported by the NationalResearch and Development Program

(NRDP) in Hungary, form a well-balanced ‘academia-industry’ cluster.They include GE Hungary Rt, a bigmanufacturing enterprise, and MT-System Ltd, an information technologySME, both of which represent industry,and the Budapest University ofTechnology and Economics, the MiskolcUniversity and SZTAKI, representingacademia.

The Figure illustrates the concept of adigital, distributed enterprise repre-senting the framework for the research.The vision of Digital Enterprises is toprovide the capability to ‘Manufacture inthe Computer’. It incorporates amodelling and simulation environmentthat enables the fabrication and/orassembly of products including the asso-ciated manufacturing processes to besimulated by the computer. This capa-bility takes into account all the variables

in the production environment, fromshop-floor processes to enterprise trans-actions. Effectively, it accommodatesthe visualisation of interacting produc-tion processes, process planning,scheduling, assembly planning, andlogistics from the resource to the enter-prise level.

The main characteristics of the projectcorrespond to the above main tenden-cies:• unified management of technical and

economical problems (includingconcurrent engineering of productsand production systems and thevertical integration of technical designand technological levels)

• purposeful, effective application ofup-to-date computerised methods tooptimise and control rapidly changingcomplex production structures in anenvironment full of uncertainties

Today’s complex manufacturing systems operate in a changing environment rifewith uncertainty. The performance of manufacturing companies ultimately hingeson their ability to rapidly adapt their production to current internal and externalcircumstances. The Hungarian national R&D project, led by SZTAKI, combinesthe concepts of intelligent manufacturing systems and digital enterprises inaddressing these problems.

Digital Factories, Production Networksby László Monostori, Géza Haidegger, József Váncza and Zsolt János Viharos

Illustration of the concept of the NRDP project on digital enterprises.



• integrated application of tele-presenceand interactive multimedia techniquestowards the establishment and indus-trial application of virtual controlrooms (VCRs), and the involvementof customers in the different phases ofdesign and production.

The project has the following — partlyoverlapping — main directions, whichwill be treated in a comprehensive way:• management and scheduling of large-

scale projects

• tele-presence and interactive multi-media

• monitoring of complex productionstructures.

This national R&D project aims to makeall the important production-relatedinformation available and manageable ina controlled, user-dependent way by theefficient application of information andcommunication technologies. The devel-opment and application of intelligentdecision support systems will help enter-

prises to cope with the problems ofuncertainty and complexity, increasetheir efficiency, join in productionnetworks and to improve the scope andquality of their customer relationshipmanagement.

Please contact:László Monostori, SZTAKITel: +36 1 279 6159E-mail: [email protected]

The GenoStar consortium was created atthe end of 1999. Partly supported by theFrench ministry of research, it bringstogether four partners: two biotechcompanies, Hybrigenics (Paris) andGENOME express (Grenoble), and tworesearch institutes, in biology, thePasteur Institute (Paris), and in computerscience, INRIA Rhône-Alpes(Grenoble). At the end of May, theconsortium presented the first version ofits platform, which will be made avail-able to academic users by the end of thisyear.

The GenoStar platform has beendesigned according to a specific view ofthe genomic world as a complex networkof biological entities and their relation-ships. As a very simple example, a geneand a protein can be seen as entitieslinked together through the ‘is-coding-for’ relationship. The gene is related toits chromosome, the chromosome to itsspecies, and so on. More abstract entitiescan be described in the same way, suchas a set of homologue genes of differentspecies or the components of a eucary-otic gene, ie the (coding) exons and the(non-coding) introns.

GenoStar is structured into fourmodules: GenoCore, GenoAnnot,GenoLink and GenoBool. As indicatedby its name, GenoCore is the centralmodule that provides the variousservices needed by the other three appli-cation modules. It is thus in charge of themanagement of the entities and theirrelationships within a knowledge baserelying on an advanced entity-relation-ship model; it offers persistence, query,and editing facilities.

Data, described as values of theattributes of these entities, must howeverbe analysed with the help of adequatemethods, which are provided by theapplication modules. Each of the threeexisting application modules,GenoAnnot, GenoLink and GenoBool,plays a different role in accordance withthe general view of the genomic worldsupported by GenoStar.

GenoAnnot allows the biologist to iden-tify regions of interest in a genomic

by François Rechenmann

As a direct consequence of the spectacular progress of experimental methodsand devices, the diversity and volume of biological data are steadily increasing.In this context, the GenoStar consortium has developed an integrated, interactiveand easy-to-use computer platform which helps the biologists to turn these datainto knowledge.

GenoStar: A Bioinformatics Platform for Exploratory Genomics

Figure 1: The results produced by several sequence analysis methods aresimultaneously displayed along the genomic sequence.



sequence. An important example of sucha region is a gene, which can be seen as aportion of the genome which containsthe information required by the cellmachinery to make a protein. In thepresent version of GenoStar, GenoAnnotprovides several methods for identifyinggenes in procaryotic (bacterial)genomes. Clearly, each time a gene isidentified, a corresponding instance ofthe gene class is created and adequatelyrelated to the other entities. The resultsof the identification methods can be

displayed on one-dimensional maps ofthe genome (see Figure 1). The biologistcan then easily compare the predictionsof concurrent methods and eventuallymake a decision on their validation.

In GenoStar, a knowledge base is popu-lated by entities and relations, which areeither imported from external databasesor produced by GenoAnnot. TheGenoLink application module allows thebiologist to explore this network of enti-ties. Complex queries can be expressed

as partial networks that are searched forin the network of the knowledge base(see Figure 2). Through this explorationprocess, the biologist may be able toinfer new relations between previouslyunrelated entities. A typical example ofsuch an inference is the prediction of thefunction of a gene (ie the function of theprotein for which the gene codes), usinginformation of other genes to which it isrelated by some pertinent relations.

The use of GenoAnnot and GenoLinkresults in the addition of instances toexisting classes: GenoAnnot createsinstances of entity classes, GenoLinkcreates instances of relationship classes.To complete the exploratory mission ofthe GenoStar platform, the GenoBoolmodule provides the user with data anal-ysis methods that can enable the identifi-cation of new classes of pertinent entitiesor relationships (Figure 3). GenoBooloffers various ways to transform databefore applying classical data analysismethods.

As stated in the introduction, progress inexperimental methods and tools isleading to the emergence of new types ofdata, which obviously require sets ofadapted methods. GenoStar has beendesigned so that it is easy not only to addmethods and strategies to the existingmodules, but also to add new modules.

At the present time, the next importantstep is to deliver a version of GenoAnnotdedicated to the analysis of eucaryoticgenomes. This is still an open problem inbioinformatics due to the existence ofvery large intergenic regions and to thestructure of the genes, which are madeup of exons and introns. This version,together with important improvementsand extensions in the other modules, willallow the consortium to deliver next yeara commercial version of the GenoStarplatform.

Link: http://www.genostar.org

Please contact: François Rechenmann, INRIATel: +33 4 76 61 53 65E-mail: [email protected]

Figure 2: A GenoLink query is a partial network of entities and relations which ismatched against the whole knowledge base.

Figure 3: GenoBool is the ‘data-mining’ module of GenoStar. Properly encoded data are submitted to data analysis methods.


EVENTS

The general chairs of this edition wereGenny Tortora and Shi-Kuo Chang,while the program chairs were FilomenaFerrucci and Giuliana Vitiello. The inter-national program committee of theconference comprised 55 representativesfrom 15 countries. The conference wasorganised by the Dipartimento diMatematica e Informatica of theUniversity of Salerno in co-operationwith ACM/SIGSOFT.

The SEKE conference is a premierforum for academic and industrialresearchers and practitioners in the fieldsof software and knowledge engineering.Continuing its long-term tradition, theaim of the conference is to provide afruitful forum of discussion aboutpresent and future issues for researchersand practitioners of Software Engineeringand Knowledge Engineering.

The main conference topics are exempli-fied by the talks given by the invitedspeakers: Lionel C. Briand illustratedwhy and how the two worlds of SoftwareEngineering and Knowledge Engi-neering can both gain by better commu-nication, Witold Pedrycz discussed thesynergies emerging between SoftwareEngineering and ComputationalIntelligence, Xindong Wu’s surveycovered the Data Mining techniques thatcan be used to distill knowledge fromdata, and Paul Dantzig described recentevolutions in the architecture and designof high volume Web sites.

A surprising number of researchersanswered to the Call for Papers. 171papers were submitted from all over theworld touching different topics of theConference. A peer review process wasapplied, each paper being reviewed bythree independent and qualified referees.As a result, the technical program wasorganised in 26 sessions and included 71

full papers, 19 short papers, and 6posters. There were sessions concernedpreeminently with the field of softwareengineering. The corresponding paperstouched classic software engineeringissues, from requirements engineering tovalidation/verification, but also newer,emerging topics that saw a large numberof contributions.

Indeed, many authors proposed solutionsto some of the new challenges posed byhypermedia and the World Wide Webwhile others addressed applications infields such as education/training orcomputer-supported cooperative work.The papers in the session on artificialintelligence approaches to software engi-neering dealt with several issues, fromproblem understanding by knowledgemodeling through the use of case-basedreasoning for domain analysis, tosupport systems for the generation ofintegration tests. There were sessionswhich provided contributions on know-ledge engineering tools and techniques,on knowledge representation andretrieval, and on soft computing. Most ofthe applied studies were featured in twosessions, one on system applications andexperience, the other on industrial appli-cations. These studies reported on theapplication of Software Engineering andKnowledge Engineering techniques,tools and processes to a specific domainor a significant software project.

Along with the regular sessions, theprogram of this edition included twothematic workshops. Both workshopsaddressed topics of current and relevantinterest — one, chaired by AthulaGinige, on Web Engineering; the other,conducted by Günther Ruhe, onSoftware Engineering Decision Support.Moreover, a half-day tutorial was deliv-ered by Colin Atkinson. This seminarpresented a model-driven (UML-based)

approach to component engineering. Theconference also hosted a Europeansession where representatives of the EUillustrated the guidelines for the prepara-tion of projects for the 6th FrameworkProgramme of the European Commission.

The conference attracted over 160participants from 23 countries and allcontinents. There were many studentsand young researchers among the partic-ipants. A relaxed and friendly atmo-sphere fostered additional ‘off therecord’ informal discussions among theparticipants, and the social programincluded a Pizza Welcome Reception, asightseeing boat trip and gala dinner, anda trip to Capri.

The proceedings of SEKE’02 werepublished by ACM Press and containpapers from the invited and contributedtalks. A special issue of the InternationalJournal of Software Engineering andKnowledge Engineering (IJSEKE) willbe dedicated to a selection of the bestpapers from the Conference. Thepurpose of the special issue, to be printedin 2003, is to offer to the authors of suchpapers an opportunity to present a moredetailed description of their work. Guesteditors of this IJSEKE special issue areFilomena Ferrucci and Giuliana Vitiello.

Information about the Conference andthe program is available athttp://www.dmi.unisa.it/seke02

Please contact:Filomena Ferrucci, Università di Salerno, ItalyTel: +39 089 965 273E-mail: [email protected]

The fourteenth International Conference on Software Engineering and KnowledgeEngineering (SEKE’02 conference), sponsored by ERCIM, was held in Sant’Angelod’Ischia, Italy, from 15 to 19 July 2002. Ten years after the 1992 conference, whichwas held in Capri, the enchanted location offered by the isle of Ischia has beenchosen to allow participants to enjoy a unique scenario for scientific discussion.

SEKE’02 Conference Reportby Filomena Ferrucci

sponsored by ERCIM


EVENTS

Trading AgentsA trading agent is a computer programthat acts on online markets on behalf ofone or more clients, trying to satisfy theirpreferences. In the trading agent competi-tion (TAC), agents are travel agents withthe goal of assembling travel packages ina five-day period, from flight tickets, tohotel rooms, and event tickets. In a game(Figure 1), eight agents, each representingeight clients, compete for a limitedamount of goods on a total of 28 differentauctions, for 12 minutes. The agentsreceive a score based on how well clientpreferences are satisfied (sum of clientutilities) and the average score over anumber of games determines the winner.

TAC combines a fairly realistic model ofthe Internet commerce of the future withchallenging problems in automatedreasoning and decision making:• Relative to their clients, TAC agents

have to solve a combinatorial assign-ment problem, where goods availableto an agent are packaged into bundles,and delivered as such to the clients.This problem is related to that of anauctioneer determining winning bidsin a combinatorial auction. Successfultrading agents must therefore makeuse of similar integer programming orconstraint programming techniques.

• Relative to other agents, TAC agentsinstead have a strategy problem. Todetermine which strategy to use in acompetitive situation is not easy.Essentially, there is no known way tocompute the best course of action. Oneimportant strategic component in TACis estimating auction closing prices.

Good software trading agents can poten-tially handle more complex combinationsof goods, larger numbers of goods andmarkets, a wider range of market types,and faster markets with more fine-grainedgoods than their human counterparts.

TAC SoftwareFor TAC-02, SICS developed open-source TAC server software, freelyavailable to all TAC competitors as wellas to other researchers, students, andeducators in the field. The previous twocompetitions, TAC-00 and TAC-01,were played on the Michigan InternetAuctionBot platform, a configurable

auction server available as an Internetservice but not available as download-able software.

The SICS TAC server (Figure 2) iswritten using a combination of SICStusProlog and Java and consists of two mainsubsystems, the Game Server, runningthe actual games and communicating

The finals of the third annual Trading Agent Competition were held on 28 July2002, co-located with AAAI-02 in Edmonton, Canada. The actual games took placeon the Internet, with the game and auction servers running at SICS in Kista,Sweden. The agents resided at the home locations of the participating researchgroups.

The Trading Agent Competition – TAC 2002by Joakim Eriksson and Sverker Janson

Figure 1: A TAC game has 8 agents, each with 8 clients, and 28 simultaneous auctions.

Figure 2: SICS TAC server and clients.


EVENTS

with trading agents, and the InformationServer, for managing participants, gamesscheduling, and collecting anddistributing game information and statis-tics via the Web and game monitoringapplets.

SICS also developed new TAC agent-ware, a Java software toolkit for devel-oping TAC agents, making it easy to getstarted. Most of the new TAC entrantsbased their agents on the SICS TACagentware.

TAC-02 Games and AgentsThe winner in this year’s tight field ofcompetitors was ‘whitebear’, developedby Ioannis A. Vetsikas at CornellUniversity with the the University ofSouthampton as a close runner-up.

The TAC-02 finals were the conclusionsof several weeks of qualification andseeding rounds involving thousands ofgames, with different combinations ofagents and different random client pref-erences:• Qualification: select top 16 out of 26

agents for the semi-finals (120games/agent)

• Seeding: seed agents into two semi-final groups (440 games/agent)

• Semi-finals: select top 4 agents ofeach group for the finals (14games/agent)

• Finals: determine winner (32games/agent).

The number of games in the finals is quitesmall, due to the desire to run them in asingle day. This increases the element ofchance in the game and is likely to bechanged in future competitions.

Each year most teams build their agentson the ideas and solutions of the previousyear’s top-performing agents. Methods,models, and technologies used in TAC-02 agents include:• Linear Programming – for finding

optimal set of travel packages, manyinspired by the optimiser used byTAC-00 winner, ATTac

• Machine Learning – automaticlearning of closing prices on hotels,trading policies for entertainment, etc

• POMDP, Partially ObservableMarkov Decision Processes

• Genetic Algorithms – for finding anoptimal set of travel packages

• Price Prediction – various ways ofpredicting price including ‘Walrasiancompetitive equilibrium’.

Many of the agent teams did not havetime to fully implement their ideas andapproaches to the TAC problem. Thereare still many untested ideas, which willguarantee interesting results for comingTAC events.

The future of TACFollowing the 2002 finals, discussionsabout the future of TAC have beenintense. The competition has increased inpopularity, and is widely recognised as aboth fruitful and entertaining means of

addressing important issues in the field ofagents. While participant performancehas increased considerably in the firstthree competitions, the TAC game is byno means a fully solved problem and willremain as the ‘TAC classic’ in futurecompetitions.

There is consensus that the TACcommunity is ready for the introductionof an additional trading game, intro-ducing new aspects and correspondingresearch issues. But there are severaldifferent candidate proposals, rangingfrom minor modifications of ‘TACclassic’ to quite different gamesinvolving spectrum auctions, stockmarkets, and supply chain management.There is also consensus that TAC will beproposed as an official workshop in amajor AI or agents conference such asIJCAI or AAMAS, potentially attractingmore interest and participants.

We look forward to seeing even moreEuropean participants in future competi-tions. You can get a head start in ‘TACclassic’ by downloading and trying outthe SICS TAC server and agentware.

Link:http://www.sics.se/tac/

Please contact:Joakim Eriksson, SICSTel: +46 8 633 1541E-mail: [email protected]://www.sics.se/~joakime/

Figure 3:Locations of the TAC-2002participants.


EVENTS

The purpose of the CONCUR confer-ences is to bring together researchers,developers and students in order toadvance the theory of concurrency andpromote its applications. Interest in thistopic is continuously growing as a conse-quence of the importance and ubiquity ofconcurrent systems and their applica-tions, and of the scientific relevance oftheir foundations. The scope covers allareas of semantics, logics and verifica-tion techniques for concurrent systems.Topics include concurrency-relatedaspects of models of computation andsemantic domains, process algebras,Petri nets, event structures, real-timesystems, hybrid systems, probabilisticsystems, model checking, verificationtechniques, refinement techniques, termand graph rewriting, distributedprogramming, logic constraint program-ming, object-oriented programming,typing systems and algorithms, security,case studies, tools and environments forprogramming and verification.

The first two CONCUR conferenceswere held in Amsterdam (NL) in 1990and 1991, with the following locationsbeing Stony Brook (USA), Hildesheim(D), Uppsala (S), Philadelphia (USA),Pisa (I), Warsaw (PL), Nice (F),Eindhoven (NL), University Park(Pennsylvania, USA), and Aalborg(DK). The proceedings of CONCUR2002 have appeared in Springer LNCSas volume 2421.

The technical programme of the confer-ence and its satellite events delivered116 presentations in total, from which 18were invited talks and invited tutorials.Out of 101 regular papers submitted toCONCUR 2002, 32 were accepted forpresentation at the conference. Theconference also included five talks byinvited speakers. Both Amir Pnueli(Weizmann Institute of Science,Rehovot, IL) and Vaughan Pratt

(Stanford University, USA) reappearedin the role of CONCUR invited speakersafter ten years, and further invited talkswere given by Wan Fokkink (CWIAmsterdam, NL), Alexander Rabinovich(Tel Aviv University, IL) and DavideSangiorgi (INRIA Sophia Antipolis, F).Additionally, there were two invitedtutorials, given by Julian Bradfield(University of Edinburgh, UK) andAndrew D. Gordon (Microsoft Research,Cambridge, UK). The best student paperaward went to ‘Local Model-CheckingGames for Fixed Point Logic with Chop’by Martin Lange.

CONCUR 2002 was accompanied by aspecial event called Tools Day organisedby Ivana âerná, and by the followingsatellite workshops:• CMCIM’02 (Categorical Methods for

Concurrency, Interaction, andMobility), organised by AlexanderKurz, held on 24 August 2002

• EXPRESS’02 (9th InternationalWorkshop on Expressiveness inConcurrency), organised by UweNestmann and Prakash Panangaden,held on 19 August 2002

• FATES’02 (Formal Approaches toTesting of Software), organised byRobert M. Hierons and Thierry Jéron,held on 24 August 2002

• FOCLASA’02 (1st InternationalWorkshop on Foundations ofCoordination), organised by AntonioBrogi and Jean-Marie Jacquet, held on24 August 2002

• INFINITY’02 (4th InternationalWorkshop on Verification of InfiniteState Systems), organised by AntonínKuãera and Richard Mayr, held on 24August 2002

• MTCS’02 (3rd International Workshopon Models for Time-Critical Systems),organised by Kim Larsen and WalterVogler, held on 24 August 2002

• PDMC’02 (1st International Workshopon Parallel and Distributed Model

Checking), organised by Lubo‰ Brimand Orna Grumberg, held on 19 August2002.

The conference took place in the CzechRepublic for the first time. The organ-isers attempted to give the conferenceadditional stimuli like moderate confer-ence fees and cheap accommodation onthe University campus to make it attrac-tive to participants from Central andEastern Europe and to young researchersall over the world.

Link:Conference web page: http://www.fi.muni.cz/concur02

Please contact:Lubos Brim, CRCIMTel: +420 54151 2323E-mail: [email protected]

The 13th International Conference on Concurrency Theory, sponsored by ERCIM,was hosted by the Faculty of Informatics of Masaryk University Brno, CzechRepublic from 19-24 August 2002. There were 174 registered participants toCONCUR 2002 and its satellite events from 18 countries around the world.

CONCUR 2002 Conference Reportby Luboš Brim

Impressions from CONCUR 2002.

sponsored by ERCIM

Pho

tos:

Jan

Hol

eãek

.

60

EVENTS

CALL FOR PAPERS

ECIR’03 — 25th EuropeanConference on InformationRetrieval ResearchPisa, Italy, 14-16 April 2003

The annual European Conference onInformation Retrieval Research,organised by the Information RetrievalSpecialist Group of the BritishComputer Society (BCS-IRSG), is themain European forum for thepresentation of research results in thefield of information retrieval.

We encourage the submission of papersreporting original, unpublished researchresults in information retrieval, broadlyintended to cover all aspects ofcomputer-based access to data withpoorly specified semantics. Thisincludes (but is not limited to) thesearching, browsing, filtering,classifying, clustering, mining, trackingand summarising of written text,hypertext, spoken text, music, images,video, and Web resources. Moreinformation on the list of topics that willbe covered by the Conference can befound on the ECIR’03 Web site.

Important Dates• Paper submission: 20 November 2002• Acceptance/rejection notification:

21 December 2002• Submission of camera-ready copy:

20 January 2003.Accepted papers will appear in theproceedings published by Springer-Verlag in the Lecture Notes inComputer Science series. Awards willbe presented to the author of the BestPaper and to the author of the BestStudent Paper.

More Information:

http://ecir03.isti.cnr.it/

ERCIM played an important role both inorganising this event (the three organ-ising bodies — the Faculty of Science ofP. J. ·afárik University, the Faculty ofMathematics, Physics and Informatics ofComenius University, and the SlovakSociety for Computer Science areconnected with ERCIM as members ofSRCIM) and in providing necessaryfunds by means of a financial grant.

The participating teams from Croatia,the Czech Republic, Germany, Hungary,Iran, Poland, Romania, Slovakia andSlovenia (with 60 contestants and teamleaders in all) demonstrated their skills inwriting and testing programs by solvingsix problems ranging from graph theoryto theory of games. The solutions wereevaluated automatically, taking intoaccount accuracy (including borderlinecases) as well as the time and spacecomplexity of the algorithms. Three goldmedals were awarded: the winner wasPeter Bella from Slovakia, and the nexttwo places were taken by Victor Costanand Daniel Dumitran from Romania.

The CEOI, which is open to secondaryschool students in Central Europe, aims to:

• encourage students’ interest in infor-matics and information technology ingeneral

• test and prove their competence insolving problems with the help ofcomputers

• promote the exchange of knowledgeand experience with other students ofsimilar interests and qualifications

• help establish personal contactsbetween young people of the CentralEuropean region.

Additionally, its purpose is to providetraining for the students participating inthe International Olympiad inInformatics (IOI), and to initiate discus-sion and cooperation in informaticseducation in the secondary schools ofCentral European countries. In our viewthis, the ninth CEOI, was successful infulfilling these aims. Next year’s CEOIwill be held in Münster, Germany, andPoland will host CEOI 2004.

Link: http://cs.science.upjs.sk/ceoi/

Please contact: Gabriela AndrejkováP. J. Šafárik University, Košice (SRCIM)Tel: +421 95 62 21 128E-mail: [email protected]

by Gabriela Andrejková and Jan Vinafi

CEOI 2002 took place from 30 June to 6 July 2002 in the Slovakian city of Kosice,under the auspices of the Ministry of Education of the Slovak Republic.

CEOI 2002 — Central European Olympiad in Informatics 2002

sponsored by ERCIM

Forthcoming Events sponsored by ERCIM

• SOFSEM 2002 — 29th Annual Conference on Current Trends in Theory and Practice of Informatics, Milovy, Czech Republic, 24-29 November 2002http://www.sofsem.cz/

• Euroweb 2002 Conference, Oxford, UK, 17-18 December 2002http://www.w3c.rl.ac.uk/Euroweb/

• Coding and Cryptography WCC2001, Versailles, France, March 24-24, 2003http://www-rocq.inria.fr/codes/WCC2003/

For more information about ERCIM’s conferences, workshops and summerschool sponsorship programme, see http://www.ercim.org/activity/sponsored.htm



EVENTS

CALL FOR PAPERS

Fourth International Workshopon Energy MinimisationMethods in Computer Visionand Pattern Recognition –EMMCVPR’2003Lisbon, Portugal, 7-9 July 2003

Many problems in computer vision andpattern recognition (CVPR) are couchedin the framework of optimisation. Theminimisation of a global quantity, oftenreferred to as the energy, forms thebulwark of most approaches in CVPR.Disparate approaches such as discreteand probabilistic formulations on the onehand, and continuous, deterministicstrategies on the other, often have opti-misation or energy minimisation as acommon theme. Instances of energyminimisation arise in Gibbs/Markovmodelling, Bayesian theory, geometricand variational approaches and in areasin CVPR such as object recognition/retrieval, image segmentation, registra-tion, reconstruction, classification anddata mining.

The aim of this workshop, the fourth in aseries, is to bring together researcherswith interests in these disparate areas,but with an underlying commitment tosome form of optimisation. Although thesubject is traditionally well representedin major conferences on CVPR, thisworkshop provides a forum whereinresearchers can report their recent workand engage in more informal discus-sions. As with the previous editions theproceedings will be published bySpringer Verlag in the Lecture Notes onComputer Science (LNCS) series.

Topics include (but are not restricted to):• Gibbs/Markov modeling • Probabilistic networks and graphical

models• Variational formulations, level sets,

and PDEs• Deformable models and registration• Graph matching• Statistical pattern recognition• Supervised learning• Unsupervised learning• VC-theory and support vector

machines

• Information theoretic methods• Model selection• Combinatorial optimisation• Interior point methods• Image reconstruction and coding• Neural networks for classification and

regression• Markov-Chain Monte Carlo methods• Relaxation labeling• Advanced mean-field methods• Self-organising networks• Evolutionary / genetic approaches• Applications.

Important Dates• Paper submission deadline:

6 January 2003• Notification of acceptance:

1 March 2003• Camera-ready paper due:

1 April 2003.

More information:http://www.emmcvpr.org/

CALL FOR PAPERS

3rd ESA Workshop on Millimetre Wave Technologyand Applications: Circuits,Systems, and MeasurementTechniquesEspoo, Finland, 21-23 May 2003

The third European Space Agency(ESA) workshop on the technology andapplications of millimetre waves will beorganised by MilliLab jointly withESA/ESTEC, VTT and HelsinkiUniversity of Technology. The aim ofthis international workshop is to bringtogether people involved in research andindustrial development of millimetrewave components and systems, and toexplore common areas and synergies inthe development of millimetre wavetechniques for commercial andscientific ground-based and space-borneapplications.

More information: http://www.vtt.fi/millilab/docs/FirstCall2003.pdf

CALL FOR PAPERS

WSCG 2003 — 11th Conferencein Central Europe on ComputerGraphics, Visualisation and Computer VisionPlzen, Czech Republic, 3-7 February 2003

the WSCG conference has been heldannually since 1992. The conference isdevoted to:• Algorithms, Parallel & Distributed

Graphics, Computer Aided GeometricDesign

• Graphics Architecture &Visualization HW, Image BasedRendering, Mathematical Aspects ofGraphics

• Global Illumination, Ray Tracing,Radiosity

• Computational Geometry, SurfaceMeshing, Modeling, ConstraintMotion, Simulation

• Virtual Reality & VR Interaction,Animation, Viewing Dynamic WorldMorphing & Warping

• Visualization, Computer Vision &Image Processing, PatternRecognition

• WWW Technologies, Hypermedia,Human Computer Interface(Graphical&Vision,Haptic)

• CAD/CAM & GIS Systems,Education Aspects, Applications.

The conferences are interdisciplinarilyoriented, presenting the state-of-the-art,as well as the latest research results andapplications. The WSCG conferencesare intended to promote the exchange ofideas over a broad range of topics.

Keynote speakers:• Gabor T. Herman: ‘Blobs for

Representing and Visualizing 3DObjects’, City University of NewYork, USA

• Dirk Bartz: ‘Large ModelVisualization: Techniques andApplications’, University ofTübingen, Germany.

More information:http://wscg.zcu.cz


EVENTS

sponsored by ERCIM

Euroweb 2002 Conference The Web and the GRID: from E-Science to E-Business Oxford, UK, 17-18 December 2002

The conference theme is intended toprompt debate on a convergence ofdevelopments pioneered for e-scienceon the GRID and as Web services toprovide business applications. Thecommunities who identify with theWeb, the GRID, Web Services, GridServices and Semantic Web should notbe isolated from each other, but need tocome together to unify their approachesto meet the real needs of information,data and knowledge technology users:• how will frameworks like Web

services, GRID Services, and .NETaddress the issue of Internet-awareprograms using the services that areoffered by other programs?

• what is XML’s role in managing androuting data and services?

• what kinds of problems lendthemselves to resource sharing?

• should Web services be inspectablewhilst running? Are there lessons tolearn from GRID services aboutquality of service management ?

• what contribution can Semantic Webtechnologies make to servicedirectories, quality of service policiesetc?

• where are the practical businessapplications requiring the terabits/secof speed, terabytes of storage andpetaflops of processing used in GRIDcomputing?

• what are the barriers to adoption andassimilation of these technologies forbusiness, and how should they beovercome?

EuroWeb 2002 will be a majorinternational forum at which research onGRIDs and Web Services is presented.EuroWeb 2002 is endorsed by the UKand Ireland office of W3C, and byIW3C2.

More information:

http://www.w3c.rl.ac.uk/Euroweb/

ERCIM News is the magazine of ERCIM.Published quarterly, the newsletter reports onjoint actions of the ERCIM partners, and aimsto reflect the contribution made by ERCIM tothe European Community in InformationTechnology. Through short articles and newsitems, it provides a forum for the exchange ofinformation between the institutes and alsowith the wider scientific community. This issuehas a circulation of over 9000 copies.

Copyright NoticeAll authors, as identified in each article, retaincopyright of their work.

ERCIM News online edition is available athttp://www.ercim.org/publication/ERCIM_News/

ERCIM News is published by ERCIM EEIG, BP 93, F-06902 Sophia-Antipolis CedexTel: +33 4 9238 5010, E-mail: [email protected] 0926-4981

Director: Bernard LarrouturouCentral Editor:

Peter [email protected]

Local Editors:AARIT: Wolfgang Hawlik

[email protected]: Martin Prime

[email protected]: Michal Haindl

[email protected]: Henk Nieland

[email protected]: Carol Peters

[email protected]: Leta Karefilaki

[email protected]: Michael Krapp

[email protected]: Patrik Hitzelberger

[email protected]: Bernard Hidoine

[email protected]: Truls Gjestland

[email protected]: Harry Rudin

[email protected]: Kersti Hedman

[email protected]: Gabriela Andrejkova

[email protected]: Erzsébet Csuhaj-Varjú

[email protected]: Ann McNamara

[email protected]: Pia-Maria Linden-Linna

[email protected]

Free subscriptionYou can subscribe to ERCIM News free ofcharge by: • sending e-mail to your local editor• contacting the ERCIM office (see address

above) • filling out the form at the ERCIM website at

http://www.ercim.org/

Call for Participation

‘iEX’ — Internet Expo 2003 Zurich, Switzerland, 5-7 February 2003

The target attendee is the Internetprofessional. The meeting has two parts:a tradeshow and a conference.Attracting some 30,000 visitors lastyear, the iEX tradeshow this year willhave three main sections:Internet/intranet solutions, networkingproducts, and Web services. Theconference portion attracted some 3000attendees last year. This year the topicsare e-commerce, intranet, Web-mastering, software development(particularly Web related), and networkplanning strategies. Workshops andcase studies are included in theprogram.

More information:

http://www.iex.ch/

Book Review

The ITU released a new publication,‘Internet for a Mobile Generation’, on 17September, 2002. If you are concerned

about themixture of twoexplosive growthareas, the Internetand mobilecommunications,you should havethis 240-pagebook at yourside. You willfind almost allthe information

you need in this book: a summary of thecompeting technologies and their appli-cations, case studies from various coun-tries, how 200 economies are performingin terms of mobile and Internet technolo-gies, regulatory and policy aspects, andsample scenarios for the future. Extractsare given on the ITU’s Web site:http://itu.int/mobileinternet/. The bookmay also be ordered through that site.This is the fourth in a series of studieswritten under the supervision of TimKelly who writes with a clear anddelightful style. Harry Rudin, Consultant and SwissLocal Editor for ERCIM News


CWI — Jaco W. de Bakker received ahigh Royal Decoration (Ridder in deOrde van de Nederlandse Leeuw),during his ‘retirement symposium’ atCWI on 30 August 2002. Jaco de Bakkeris one of the pioneers of computerscience in The Netherlands, especially inthe field of mathematical semantics ofprogramming languages and reasoningon program correctness. He worked for38 years at CWI, and put the word‘Informatics’ in the name of the formerMathematical Centre. From 1973 on, DeBakker was Professor at the VrijeUniversiteit in Amsterdam, and in 1989he became a Member of the RoyalNetherlands Academy of Arts andSciences. He wrote more than 150 scien-tific articles and books. De Bakker isproud that 32 scientists who worked inhis group were appointed professor. Alsoat managerial level, De Bakker played animportant role. CWI will miss him andwishes him all the best for the future.

CWI — Professor Edsger WybeDijkstra, world-famous pioneer of thescience and industry of computing, diedafter a long struggle with cancer on 6August 2002 at his home in Nuenen, theNetherlands, at the age of 72. Dijkstraworked at the Mathematisch Centrum(MC/CWI) between 1952 and 1962.During his Amsterdam years, he workedas the first ‘programmer’ (his ownmodest description) in the Netherlands,participating in the design of the MC’sfirst commercially marketed computer,

the X1. Perhaps his greatest achievementduring these years was the writing withJaap Zonneveld of the world’s firstALGOL60 compiler. From 1962 until1984 he held a chair at the EindhovenUniversity of Technology. His fame wasaugmented through his fundamentalstudies of parallel programming and hisinsights into the construction of correctprograms, and he was an eloquent advo-cate of the methodology of structuredprogramming. Dijkstra was a BurroughsCorporation research fellow from 1973-1984. From 1984 until his retirement in1999 he worked at the University ofTexas in Austin. He wrote over 1300books and papers, all of which are digi-tally accessible. Dijkstra was the 1972recipient of the ACM Turing Award,often viewed as the Nobel Prize forcomputing. He was also a recipient ofnumerous honorary degrees and awards.Dijkstra is universally acclaimed as themost famous scientist in the history ofthe Mathematical Centre, and wasarguably one of the most influentialcomputer scientists of the twentiethcentury. Dijkstra is survived by his wife,Ria, and three children. For more infor-mation, see: http://www.cs.utexas.edu/users/EWD/

CNR — Piero Maestrini has beenappointed Director of the newly consti-tuted ‘Istituto di Scienza e Tecnologiedell’Informazione’ (ISTI), located in theCNR Research Area, Pisa. ISTI-CNR isthe result of a union between the CNR

Institutes of CNUCE and IEI and willconduct R&D activities in the fields ofcomputer science and IT. Currently, themain areas of interest are currently: soft-ware engineering, formal methods forspecification and verification, networkedmultimedia information systems, Web-based information systems, knowledgediscovery and data mining, image andsignal processing, visual computing andcomputer graphics, human-computerinteraction, high performancecomputing, wireless and mobilenetworks, dependable computing, struc-tural mechanics and spaceflightdynamics. Further information can befound at http://www.isti.cnr.it/.

IN BRIEF

ERCIM PhD Fellowship Programme

ERCIM offers 18-month fellowships in leading European information technologyresearch centres. Fellowships are available in areas covered by the ERCIMWorking Groups and other fields. This offer is available for PhD holders from allover the world.

Deadline for application: 31 October 2002

Conditions Applicants must: • have a PhD degree (or equivalent) or be in the last year of the thesis work• be fluent in English • be discharged or get deferment from military service.

Fellowships are of 18 months duration, generally spent in two institutes. The fellow willreceive a monthly allowance which may vary depending on the country. In order toencourage the mobility, a member institution will not be eligible to host a candidate of thesame nationality.

Detailed information and the online application form is available at: http://www.ercim.org/fellowship/

Pho

to C

WI

ERCIM – The European Research Consortium for Informatics and Mathematics is an organisationdedicated to the advancement of European research and development, in information technologyand applied mathematics. Its national member institutions aim to foster collaborative work withinthe European research community and to increase co-operation with European industry.

Order Form

If you wish to subscribe to ERCIM News free of charge

or if you know of a colleague who would like toreceive regular copies of

ERCIM News, please fill in this form and wewill add you/them to the mailing list.

Send, fax or email this form to:ERCIM NEWS

Domaine de VoluceauRocquencourt

BP 105F-78153 Le Chesnay Cedex

Fax: +33 1 3963 5052 E-mail: [email protected]

Data from this form will be held on a computer database. By giving your email address, you allow ERCIM to send you email

Name:

Organisation/Company:

Address:

Post Code:

City:

Country

E-mail:

✄

Fraunhofer-Gesellschaft, Information and Communication Technology AllianceSchloß Birlinghoven, D-53754 Sankt Augustin, GermanyTel: +49 2241 14 0, Fax: +49 2241 14 2889http://www.fraunhofer.de/

Consiglio Nazionale delle Ricerche, ISTI-CNRArea della Ricerca CNR di Pisa, Via G. Moruzzi 1, 56124 Pisa, ItalyTel: +39 050 315 2878, Fax: +39 050 315 2810http://www.isti.cnr.it/

Centrum voor Wiskunde en InformaticaKruislaan 413, NL-1098 SJ Amsterdam, The NetherlandsTel: +31 20 592 9333, Fax: +31 20 592 4199http://www.cwi.nl/

Institut National de Recherche en Informatique et en AutomatiqueB.P. 105, F-78153 Le Chesnay, FranceTel: +33 1 3963 5511, Fax: +33 1 3963 5330http://www.inria.fr/

Central Laboratory of the Research CouncilsRutherford Appleton LaboratoryChilton, Didcot, Oxfordshire OX11 0QX, United KingdomTel: +44 1235 82 1900, Fax: +44 1235 44 5385http://www.cclrc.ac.uk/

Swedish Institute of Computer ScienceBox 1263, SE-164 29 Kista, SwedenTel: +46 8 633 1500, Fax: +46 8 751 72 30http://www.sics.se/

Technical Research Centre of FinlandP.O. Box 1200FIN-02044 VTT, FinlandTel:+358 9 456 6041, Fax :+358 9 456 6027http://www.vtt.fi/tte

Foundation for Research and Technology – HellasInstitute of Computer ScienceP.O. Box 1385, GR-71110 Heraklion, Crete, GreeceTel: +30 81 39 16 00, Fax: +30 81 39 16 01http://www.ics.forth.gr/FORTH

Czech Research Consortium for Informatics and MathematicsFI MU, Botanicka 68a, CZ-602 00 Brno, Czech RepublicTel: +420 2 688 4669, Fax: +420 2 688 4903http://www.utia.cas.cz/CRCIM/home.html

Swiss Association for Research in Information TechnologyDept. Informatik, ETH-Zentrum, CH-8092 Zürich,SwitzerlandTel: +41 1 632 72 41, Fax: +41 1 632 11 72http://www.sarit.ch/

Slovak Research Consortium for Informatics andMathematics, Comenius University, Dept.of ComputerScience, Mlynska Dolina M, SK-84248 Bratislava,Slovakia, Tel: +421 7 726635, Fax: +421 7 727041http://www.srcim.sk

Magyar Tudományos AkadémiaSzámítástechnikai és Automatizálási Kutató IntézeteP.O. Box 63, H-1518 Budapest, HungaryTel: +36 1 279 6000, Fax: + 36 1 466 7503http://www.sztaki.hu/

Trinity CollegeDepartment of Computer Science,Dublin 2, IrelandTel: +353 1 608 1765, Fax: 353 1 677 2204http://www.cs.tcd.ie/ERCIM

Austrian Association for Research in ITc/o Österreichische Computer GesellschaftWollzeile 1-3, A-1010 Wien, AustriaTel: +43 1 512 02 35 0, Fax: +43 1 512 02 35 9http://www.aarit.at/

Norwegian University of Science and Technology Faculty of Information Technology, Mathematics andElectrical Engineering, N 7491 Trondheim, NorwayTel: +47 73 59 80 35, Fax: +47 73 59 36 28http://www.ntnu.no/

Fonds National de la Recherche20, Montee de la PetrusseL-2912 LuxembourgTel. +352 26 1925 31, Fax +352 26 1925 35http:// www.fnr.lu

You can also subscribe to ERCIM News and order back copies by filling out the form at the ERCIM website at http://www.ercim.org/publication/Ercim_News/

WEB SEMÀNTICA

Documents

semantic weband

semantic webby terry

semantic webby michael

semantic webby sofia

semantic webby lynda

semantic webby ying

semantic webby santtu

semantic webby benjamin