NEWSLETTER News from the BelgiumNetherlands Association for Artificial Intelligence
August 2002 Vol. 19, No. 4 ISSN 1566-8266
BNAIC 2002 programme The Horowitz factor Removing ‘Volkert’ from court decisions
AI Considered Harmless Editor-in-chief On August 6, 2002, the most famous Dutch computer scientist Edsger Dijkstra died at his home in Nuenen. Dijkstra made a rich variety of contributions to computer science and was one of the founding fathers of Dutch computer science. The majority of his writings are collected in 1317 numbered notes (called “EWDs”). Dijkstra wrote eloquently and expressed his opinions unabashed; your editor-in-chief compiled a selection of AI-related quotations from the EWDs. Actually, Dijkstra was not very fond of any interdisciplinary branch of computer science, and AI was on of them: Usually I don't need to talk about artificial intelligence as it is only a specific area of potential machine application and as such outside the scope of computing science proper. [on using AI for programming:] The problem with the “smart” machines is the same as we had with all the programming language “features”: each new layer of “user-friendliness” blurs the specification and thereby makes the system more risky to use. Science fiction and science reality in computing, EWD 952 (1986) It is not only the food industry that employs “natural” as a vague recommendation. (The almonds the airline serve with the drinks are always “Natural hickory flavored”.) In the worlds of logic and computing we find that same laudatory undertone in “natural deduction” and “natural language (programming)”. On closer scrutiny it is a euphemism for something like “appealing to the uneducated”. A bagatelle for the left hand, EWD 800 (1982) In EWD 899 Dijkstra wrote a devastating review of a paper on default reasoning. The review begins with: For someone without much prior knowledge to the material in question, the paper is quite revealing as it gives some idea of the stuff so-called “expert-systems” are made of. Starting from half-truths and treating them as the whole truth, contradictions are easily reached. A review of a paper in AI, EWD 899 (1984) The concept of software agents was probably not to Dijkstra’s liking: Is anthropomorphic thinking bad? Well, it is certainly no good in the sense that it does not help. [...] But is it also bad? Yes it is [...] the anthropomorphism becomes more misleading than helpful. On anthropomorphism in science, EWD 936 (1985) Dijkstra was not easily seduced in a discussion on thinking computers: Ik zal het niet meer meemaken, maar als het een beetje meezit is over honderd jaar de vraag of computers kunnen denken niet meer dan een historisch curiosum uit de twintigste eeuw. Hoe onbelangrijk het is of duikboten kunnen zwemmen, EWD 1056 (1989) John von Neumann speculated about computers and the human brain in analogies sufficiently wild to be worthy of a medieval thinker and Alan M. Turing thought about criteria to settle the question of wheather Machines Can Think, a question of which we now know that is about as relevant as the question of whether Submarines Can Swim. The threats to computing science, EWD 898 (1984) As becomes clear from the above, AI was not Dijkstra’s cup of tea. Nevertheless, his work has been seminal for the development of computer science, and as such, also for the development of AI. We can be proud that our computer science community has brought forth such a great mind. EWD Archive: http://www.cs.utexas.edu/users/EWD/ VPRO Interview: http://www.vpro.nl/wetenschap/index.shtml?3626936+4257491+7958499
BNVKI Newsletter
78
August 2002
TABLE OF CONTENTS AI Considered Harmless (Editor-in-Chief) ........................................................................................................... 78 Table of Contents .................................................................................................................................................. 79 BNVKI-Board News (Joost Kok).......................................................................................................................... 80 Groningen Wins at Open Dutch Student Championship Computer-RoShamBo (Jeroen Donkers) ...................... 80 ECAI 2002: In Search of the Horowitz Factor (Bart Jansen) ................................................................................ 80 The Way to AI (Jaap van den Herik)..................................................................................................................... 82 Announcements .................................................................................................................................................... 82 BNAIC’02 Programme .................................................................................................................................... 82 Call for Papers CLIN 2002 .............................................................................................................................. 87 Call for Papers Twelfth Dutch-Belgian Conference on Machine Learning .................................................... 87 Subsidie Programma Neurale Technologie...................................................................................................... 88 Section AI Education (Evert van de Vrie)............................................................................................................. 90 SIKS (Richard Starmans) ...................................................................................................................................... 90 SIKS Masterclass on Human Computer Interaction ........................................................................................ 90 SIKS Day 2002 ................................................................................................................................................ 92 SIKS Masterclass on Security and Privacy in Cyberspace .............................................................................. 92 Dutch-Belgian Information Retrieval Workshop ............................................................................................. 93 Section Knowledge Systems in Law and Computer Science (Sien Moens).......................................................... 93 Feasibility of Reliable (Semi)-Automatic Anonymizing of Court Decisions (Marie-Francine Moens) .......... 93 Wetgevende Procedures en de Kwaliteit van Wetgeving in een Automatiseringscontext (Arno Lodder)....... 95 Conferences, Symposia, Workshops ..................................................................................................................... 96 E-mail addresses Board Members/ Editors BNVKI Newsletter / How to become a member?/ Submissions ....... 99
The BNVKI is sponsored by BOLESIAN
BNVKI Newsletter
79
August 2002
(tUL). Next to the 9 competitors, 13 standard bots joined the competition. Moreover, the world champion from 1999, IOCAINE POWDER by Dan Egnor played in the tournament with a Java version. Of course, this was announced in advance to the competitors. Each match lasted 1000 moves and 500 complete rounds were held. It meant that every program made 22 × 500 × 1000 moves. The outcome of the tournament was the average score over all rounds.
BNVKI-Board News Joost Kok I am writing in the holiday period. In fact, I am typing from an Internet connection in Terschelling (which was quite hard to find). In the Dutch press Terschelling obtained a bad name: at least 9 teenagers had drunk too much alcohol and had to be transported to the main land, but from where I am now it looks quite peaceful (also a bit rainy).
RESULTS
All the main sport events of the summer, like the Tour de France, Wimbledon, Robocup and Skûtsjesilen, are almost over. One of the major sport events, sponsored by the BNVKI, was the 7th Computer Olympiad in Maastricht, including the 10th World Computer Chess Championship. Below you will find a report on a particular branch: computer-Roshambo. A general report on the Olympiad will be published in the coming issue.
1. 2. 3. 4. 5. 6.
For me the summer is not yet over: next week I go to Legoland and continue to the PKDD conference in Helsinki and a bioinformatics school in Patras. Only after these events the teaching season starts again. Luckily, I can look forward to a trip to Leuven later this year. Don’t forget to registrate!
7. 8. 9.
IOCAINE POWDER got a score of 7398, implying that URZA from Groningen played better than the 1999 world champion!
Groningen Wins at Open Dutch Student Championship Computer-Roshambo
For more information on this competition, see http://www.cs.unimaas.nl/~donkers/games/roshambo
Jeroen Donkers IKAT, Universiteit Maastricht
ECAI2002: In Search of the Horowitz Factor
During the 7th Computer Olympiad in Maastricht (5-11 July 2002), a programming competition was held for Dutch and Belgian students. The task was to write a Java program that could play the game of Roshambo. This game, also called Paper-scissorsrock, is well known all over the world. Two players sit face-to-face. They count to three and then simultaneously make a symbol with their right hand: a fist (rock), flat hand (paper) or two fingers spread out (scissors). Now rock wins from scissors, scissors win from paper and paper wins from rock. Equal symbols is a draw. The optimal strategy in this game on the long term is to play randomly. In a tournament, however, time is limited and nonoptimal opponents are present. To win the tournament, a player should predict the opponents behaviour and be unpredictable itself.
Bart Jansen ARTI, Vrije Universiteit Brussel As a starting researcher I had the opportunity to visit the ECAI conference in Lyon (July 21-26, 2002). At this early stage in my own research, I did not present a personal contribution; instead, I enjoyed a wide range of talks, met interesting people and drew a few conclusions for later conferences. I much enjoyed the fact that some speakers got very much into technical details, while others only discussed their findings on a very high, almost philosophical level. However, I quite often noticed a correlation between detail of the talk and language skills of the speaker: good speakers discussed high level issues, others stressed the technical details. In a few cases, a high level talk got so high level that
In the competition, 9 teams participated from universities of Groningen (RUG), Delft (TUD), Amsterdam (UvA) and Maastricht/Diepenbeek
BNVKI Newsletter
URZA (8903) Martijn Muurman and Enno Peters, RUG PLAYERV110 (7319) Steven de Rooij, Dave van Soest, and Simon Pauw, UvA MARSHALBOT (5932) Vincent de Boer, TU Delft HAVOC (5216) Georgios Sarigiannidis and Aqab Bin Talal, TU Delft KICKINBACK (4994) Peter Bakema, RUG ROSHAMBIERTJE (4839) Joost Vanstreels and Rogier Vos, tUL DEAN (4595) Mathijs de Boer, RUG VLUGGERTJE (4348) Koen Vossen, tUL TEAM47 (-480) Petra Coenegrachts, Bart Moelens and Eva Westaedt, tUL
80
August 2002
audience could compare itself the difference between pure synthesized music and music played by a computer that learned these variations. Interestingly, individual characteristic differences could be identified as well, by simply comparing the amplitude and tempo space of different performers. As Widmer suggested, the unconventional curves in this space observed during analysis of performances by Horowitz might be the so-called Horowitz factor.
the message learnt from the talk was quite obvious. For instance Pat Hayes opened the conference as an invited speaker with a talk on the semantic web. Being a very good speaker, he could hardly be stopped. The talk was appreciated very much. On hindsight, information density in the talk seemed quite low: Hayes argued that instead of inventing new ways to represent content and meaning, we should accept the existence of different formalisms and study how a server can negotiate with applications in which form the content is delivered. This simple message contrasted with speakers who had a contribution to their field to present and got only 25 minutes or less.
ROBOTS AND NEURAL NETWORKS Sebastian Thrun presented his work on probabilistic robotics, and terminated his talk with nine open research issues for the field of robotics. Some questions were related to map building: how to obtain better maps that are consistent with the laws of geometry, physics, and with common sense? Other questions were related to probabilistic control: how to define performance functions, how to extend algorithms for a non-Markov world, how to build non-monotonic robots. A single question concerned a dream of Thrun: how can programming code be used as domain knowledge for problem solving?
LEARNING TO PLAY MUSIC It seemed that most passionate talks were given by the invited speakers Sebastian Thrun, Pieter Adriaans and Gerhard Widmer. Especially the two talks by Gerhard Widmer had a very high 'Horowitz' factor to my personal opinion, even without the promised presence of Horowitz, the Ukrainin pianist. Using machine learning techniques and data mining methods Widmer investigated how musicians put expression in their musical performance. He compared performances by highly qualified musicians with the original sheet music. As could be expected, no performer gave an exact rendition of the sheet music. All musicians seemed to stray from the written music in a systematic manner, resulting in interesting information for the field of musicology. The
Some speakers managed to present their ideas or findings in a very clear and accessible way. For instance Jacek Mandziuk from the Warsaw University of Technology reported on experiments comparing the time series prediction capabilities of several network architectures. Surprisingly, feedforward networks performed better than recurrent networks, which were actually believed to be better suited for this task. Comparison results and a first analysis of the cause of this phenomenon were presented. VISION FOR TRAFFIC MONITORING My own research interests lie in visual event recognition, and in that area there was the group of professor Nagel presented two papers on their cognitive vision system for road traffic monitoring. The system was already capable of detecting an important set of traffic events. Professor Nagel reported on how an internal representation of the temporal developments within the recorded scene can be obtained. The system is based on a fuzzy metric-temporal Horn logic. He explained how the system now is tuned to represent groups of cars, for instance queues instead of individual cars. The group is not aiming at real-time behaviour, but instead wants superior monitoring quality. Ralf Gerber reported on an extension with a natural language component, such that textual descriptions can be obtained from the visual input. He is also experimenting with the reverse operation:
Gerhard Widmer.
BNVKI Newsletter
81
August 2002
2002/2003 the assessment of the domain of Computer Science research will take place by a VSNU committee. Within this framework the computer-science oriented AI groups will be assessed too. With much commitment the BNVKI looks forward to the outcome of both assessments.
constructing a visual representation from the textual descriptions. He wants to extend the system such that it learns from the differences between the original video input and reconstructed animation of it. LESSONS LEARNT
In this issue there are no reviews of Ph.D. theses. We would like to encourage our readers and in particular the promotores to stimulate contributions reviewing theses that are defended.
During my first conference, I understood that you can't estimate the value of new work simply by attending the presentation. Only by knowing previous work and studying the papers, a clear insight can be obtained. In order to avoid deceptions, one should first briefly check the proceedings and perhaps even other papers before attending the talks. Something I definitely needn't tell to an experienced scientist.
H.S. Reijers (August 28, 2002). Design and Control of Workflow Processes Business Process: Management for the Service Industry. TU Eindhoven. Promotores: Prof.dr. K.M. van Hee and Prof.dr.ir. W.M.P. van Aalst. W.E. de Paepe (September 30, 2002). Complexity Results and Competitive Analysis for Vehicle Routing Problems. TU Eindhoven. Promotores: Prof.dr. J.K. Lenstra and Prof.dr. A.G. de Kok.
The Way to AI Jaap van den Herik IKAT, Universiteit Maastricht
M. Lamers (October 4, 2002). Neural Networks for Analysis of Data in Environmental Epidemiology. Universiteit Leiden. Promotor: Prof.dr. J.N. Kok, co-promotor: Dr. E. Lebret.
The way to Artificial Intelligence starts at cognition, reasoning and intelligent behaviour. After some time the inquisitive traveller stops at Mathematics and Logic. These two disciplines are exact, precise and attract all kind of intelligence. For this reason we have included in our list of announcements a Ph.D. thesis on Vehicle Routing Problems.
W.C.A. Wijngaards (October 14, 2002). Agent Based Modelling of Dynamics: Biological and Organisational Applications. VU Amsterdam. Promotor Prof.dr. J. Treur, co-promotor Dr. C.M. Jonker.
The start of a new academic year coincides with appointment of many fresh AiOs, of whom we expect a thesis after four years. Following this line of reasoning we would expect a long list by now, consisting of all AiOs appointed four years ago. This turned out not to be the case, since most AiOs prefer to have their Ph.D. defence before the summer holidays instead of thereafter.
R. Serban (November 5, 2002). The Private Cyberspace Modeling Electronic Environments inhabited by Privacy-concerned Agents. VU Amsterdam. Promotor: Prof.dr. R.P. van de Riet.
ANNOUNCEMENTS
Statistics teach us that the AiO path, i.e., the way to artificial intelligence, usually takes four years or somewhat more. This implies that the next list of announcements will be longer than the current one, which contains only five Ph.D. theses. All five promovendi are congratulated with the publication of their thesis. This holds true for promotores, copromotores and universities, too.
BNAIC’02 Programme October 21–22, 2002, Leuven The 14th Belgian-Dutch Conference on Artificial Intelligence is organized by the Katholieke Universiteit Leuven, the Université Libre de Bruxelles and PharmaDM under auspices of BNVKI/AIABN (Belgian-Dutch Association for AI) and SIKS (School for Information and Knowledge Systems). BNAIC is an annual conference and will be held on Monday October 21 and Tuesday October 22, 2002 at the Faculty Club, Groot Begijnhof in Leuven.
Starting in September 2002 there will be two assessment committees for which Ph.D. theses belong to the countable part of their work. First, we have the assessment of the Research School SIKS, which takes place on September 16, 17 and 18. This assessment is important for SIKS’s renewed recognition for a period of five years by the KNAW committee, the ECOS. Second, in the years BNVKI Newsletter
82
August 2002
MONDAY, OCTOBER 21 09.00 – 09.45 09.45 – 10.00 10.00 – 11.00 11.00 – 11.20 11.20 – 12.20
Registration Opening Invited talk: Hugues Bersini, Université Libre de Bruxelles Coffee Break Paper presentations: Session 1
•
Reviewing the Design of DAML+OIL: an Ontology Language for the Semantic Web Heiner Stuckenschmidt Approximate Information Filtering with Multiple Classification Hierarchies
Session 2B: Logic • Dagmar Provijn How to obtain elegant Fitch-style proofs from Goal directed ones • Marc Denecker, Nikolay Pelov & Maurice Bruynooghe Ultimate Well-founded and Stable Semantics for Logic Programs with Aggregates • Alexander Serebrenik & Danny De Schreye On termination of meta-programs • Henry Prakken An exercise in formalising teleological casebased reasoning
Session 1A: Bayesian Reasoning I • Marco A. Wiering Hierarchical Mixtures of Naive Bayesian Classifiers • Eveline M. Helsper & Linda C. van der Gaag Building Bayesian Networks through Ontologies • Wim Wiegerinck & Tom Heskes IPF for discrete chain factor graphs Session 1B: Language • Antal van den Bosch & Sabine Buchholz Shallow parsing on the basis of words only: A case study • Joachim De Beule, Joris Van Looveren & Willem Zuidema From perception to language: grounding formal syntax in an almost real world • Gabriel G. Infante-Lopez, Maarten de Rijke & Khalil Sima’an A General Probabilistic Model for Dependency Parsing
Session 2C: Intelligent User Interfaces • Eveliene de Vos, Cilia Witteman & RobbertJan Beun Embodied Conversational Agents in HumanComputer Interaction • P.P.A. Storms, E.J. Herweijer & C.J. van Aart Experimentally Established Design Guidelines for Embodied Conversational Agents • Vojtech Knezu & Leon J.M.Rothkrantz A System for Automated Bookmark Management • R.M. Schaar, L.J.M. Rothkrantz, M. Lassche & M.V. Jonkers Agent-Based Intelligent Personal Unified Messaging
Session 1C: Qualitative Reasoning • Paulo Salles, Bert Bredeweg, Symone Araujo & Walter Neto Qualitative Models of Interactions Between Two Populations • Niels Peek Representation of decision-theoretic plans as sets of symbolic decision rules • Silja Renooij, Simon Parsons & Pauline Pardieck Using Kappas as Indicators of Strength in QPNs
14.40 – 15.30 15.30 – 15.50 15.50 – 16.50
16.50 – 18.05 12.20 – 13.20 13.20 – 14.40
Lunch Paper presentations: Session 2
Demo Session • Evert van de Vrie LOK: Implementation of a platform for distributed development and use of educational tasks • Michael Kentrop & Frank van Dijk Approach • Frances Brazier, David Mobach, Benno Overeinder, Sander van Splunter, Maarten van Steen & Niek Wijngaards AgentScape Demonstration
Session 2A: KBS & Ontologies • Christof Monz & Maarten de Rijke Knowledge-Intensive Question Answering • Henk-Jan Lebbink, Cilia L.M. Witteman & John-Jules Ch. Meyer Ontology-Based Knowledge Acquisition for Knowledge Systems • Ian Horrocks, Peter F. Patel-Schneider & Frank van Harmelen
BNVKI Newsletter
Short presentations of demos and posters Coffee Break Invited talk: Nick Jennings, University of Southampton Automated Haggling: Building Artificial Negotiators Demo and Poster Sessions
83
August 2002
M. Dastani, F. de Boer, F. Dignum, W. van der Hoek, M. Kroese & J. J. Meyer Implementing Cognitive Agents in 3APL Carlos Areces & Juan Heguiabehere HyLoRes: A hybrid logic prover based on directresolution Leo Hermans Knowledge Intensive Content Model Management Within Integrated Back offices N.W.M. Schoot & W. Jansweijer Improving the quality of information in document based communications using a reusable multi-agent system.
•
Poster Session • Mehdi Dastani, Virginia Dignum & Frank Dignum Organizations and Normative Agents • Wojciech Jamroga Multiple Models of Reality and How to Use Them • Herman ter Horst, Mark van Doorn, Natasha Kravtsova, Warner ten Kate & Daniel Siahaan Context-aware Music Selection Using Knowledge on the Semantic Web • H.H.L.M. Donkers, J.W.H.M. Uiterwijk & H.J. van den Herik Learning Opponent-Type Probabilities for PrOM Search • R.P.J. van der Krogt, L.D. Aronson, N. Roos, C. Witteveen & J. Zutt Tactical Planning using Heuristics • Mehdi Dastani & Leon van der Torre An Extension of BDICTL with Functions and Components • Bert Van Nuffelen Reasoning with preferences in ID-Logic • Christiaan Fluit, Marta Sabou & Frank van Harmelen Ontology-based Information Visualisation • M. Apistola, F.M.T. Brazier, O. Kubbe, A. Oskamp, M.H.M. Schellekens & M.B. Voulon Legal aspects of agent technology • Jaap Kamps & Maarten Marx Words with Attitude • David Gilis & Marc Denecker Compositionality Results for Stratified Nonmonotone Operators • Siegfried Nijssen & Joost Kok Tree Sets: Towards a Set-Oriented View on Multi-Relational Data Mining • Jan Struyf, Jan Ramon & Hendrik Blockeel Compact representation of knowledge bases in ILP • Mar Marcos, Hugo Roomans, Annette ten Teije & Frank van Harmelen Improving medical protocols through formalisation: a case study
•
• • • •
BNVKI Newsletter
•
•
•
•
• • •
• •
• • •
• • •
84
Peter van der Putten, Martijn Ramaekers, Marten den Uyl & Joost Kok A Process Model for a Data Fusion Factory M.M. Drugan, D. Thierens & L.C. van der Gaag MDL-based Feature Selection for Bayesian Network Classifiers T. Lenaerts, A. Defaweux, P. van Remortel & B. Manderick Multi-level Selection in a Simple Evolutionary Model Alexander Ypma & Tom Heskes Clustering web surfers with mixtures of hidden Markov models M.H. ter Brugge, J.A.G. Nijhuis & L. Spaanenburg Morphological Template Decomposition for DT-CNN Piet van Remortel, Tom Lenaerts & Bernard Manderick Testing the Overall Functional Robustness of 2D CA Phenotypes for Development E.M. Oost, S.H.G. ten Hagen & F.H. Schulze Extracting multivariate power functions from complex data sets. Veska Noncheva & Nuno Cavalhiero Marques Agent’s Belief: A Stochastic Approach Arjen Vollebregt, Henk Hesselink, Daan Hannessen & Jelle Beetstra Modelling Crew Assistants with Multi-Agent Systems in Aircraft Tom Heskes & Onno Zoeter Expectation propagation for approximate inference in dynamic Bayesian networks Jan van den Berg, Uzay Kaymak & WillemMax van den Bergh Fuzzy Classification by Using ProbabilityBased Rule Weighting Stefan Kleijkers, Floris Wiesman & Nico Roos A Mobile Multi-Agent System for Distributed Computing Edwin D. de Jong & Tim Oates A Coevolutionary Approach to Representation Development Raymond Kosala, Jan Van den Bussche, Maurice Bruynooghe & Hendrik Blockeel Information Extraction in Structured Documents using Tree Automata Induction Paul E.M. Huygen Use of Bayesian Belief Networks in legal reasoning Kurt Schelfthout & Tom Holvoet “To do or not to do”: The Individual’s Model for Emergent Task Allocation Berend Jan van der Zwaag, Kees Slump & Lambert Spaanenburg Process Identification Through Modular Neural Networks and Rule Extraction August 2002
• • •
• •
• •
•
•
Alexander Serebrenik & Danny De Schreye Inference of termination conditions for numerical loops Frances Brazier, Benno Overeinder, Maarten van Steen & Niek Wijngaards Generative Migration of Agents C.M. Jonker, A.P.G. de Kock, J. Meijer & B.J. Vermeulen Deliberate Evolution Agents: Comparing Reproduction Strategies Catholijn M. Jonker & Jan Treur A Dynamic Perspective on an Agent’s Mental States and Interaction with its Environment Catholijn M. Jonker, Jacky L. Snoep, Jan Treur, Hans V. Westerhoff & Wouter C.A. Wijngaards BDI-Modelling of Intracellular Dynamics Catholijn M. Jonker & Jan Treur Analysis of the Dynamics of Reasoning Using Multiple Representations Catholijn M. Jonker, Jan Treur & Wouter C.A. Wijngaards Temporal Languages for Simulation and Analysis of the Dynamics Within an Organisation Catholijn M. Jonker, Jan Treur & Wouter C.A. Wijngaards Requirements Specification and Automated Evaluation of Dynamic Properties of a Component-Based Design
Robert E. Keller, Walter A. Kosters, Martijn van der Vaart & Martijn D.J. Witsenburg Genetic Programming Produces Strategies for Agents in a Dynamic Environment
Session 3C: Data Mining I • Birgit Hay, Geert Wets & Koen Vanhoof Web Usage Mining by means of Multidimensional Sequence Alignment Methods • Jeannette M. de Graaf, Walter A. Kosters, Wim Pijls & Viara Popova A Theoretical and Practical Comparison of Depth First and FP-growth Implementations of Apriori 10.20 – 10.40 10.40 – 12.20
Coffee Break Paper presentations: Session 4
Session 4A: Multi-Agent Systems • Jeroen Valk, Cees Witteveen Multi-Agent Coordination in Planning • Nico Roos, Annette ten Teije, André Bos & Cees Witteveen Multi-Agent Diagnosis with spatially distributed knowledge • J. Zutt, L.D. Aronson, R.P.J. van der Krogt, N. Roos & C. Witteveen Multi-Agent Transport Planning • Hans Tonino, André Bos, Mathijs de Weerdt & Cees Witteveen Plan Coordination by Revision in Collective Agent Based Systems • Niek Wijngaards, Benno Overeinder, Maarten van Steen & Frances Brazier Supporting Internet-Scale Multi-Agent Systems
18.30 – 20.15 Reception at Town Hall 20.15 – 23.00 Conference dinner TUESDAY, 22 OCTOBER 08.40 – 09.40
09.40 – 10.20
Invited talk: Hector Levesque, University of Toronto Cognitive Robotics at the University of Toronto Paper presentations: Session 3
Session 3A: Fuzzy Reasoning • Jan van den Berg, Uzay Kaymak & WillemMax van den Bergh Probabilistic Reasoning in Fuzzy Rule-Based Systems • Jeroen Eggermont Evolving Fuzzy Decision Trees for Data Classification Session 3B: Evolutionary Computing I • Karl Tuyls, Tom Lenaerts, Katja Verbeeck, Sam Maes & Bernard Manderick Towards a Relation Between Learning Agents and Evolutionary Dynamics
BNVKI Newsletter
85
August 2002
Session 5B: Evolutionary Computing II • Yiu-Fai Cheung & L.J.M. Rothkrantz Broadcast Information Topic Segmentation BITS • Peter A.N. Bosman & Dirk Thierens Multi–objective optimization with diversity preserving mixture–based iterated density estimation evolutionary algorithms • Siegfried Nijssen & Thomas Bäck An Analysis of the Behaviour of Simplified Evolutionary Algorithms on Trap Functions • W.B. Langdon Size of Random Programs to ensure Uniformity • Jeroen Eggermont & Tom Lenaerts Dynamic Optimization using Evolutionary Algorithms with a Case-based Memory
Session 4B: Applications • Pim van Leeuwen, Henk Hesselink & Jos Rohling Scheduling Aircraft Using Constraint Satisfaction • Laura Maruster, Ton Weijters, Geerhard de Vries, Antal van den Bosch & Walter Daelemans Logistic-Based Patient Grouping for Multidisciplinary Treatment • Frans Voorbraak Uncertainty in AI and Bioinformatics • Radboud Winkels, Alexander Boer & Rinke Hoekstra Lessons Learned in Legal Information Serving • Chris van Aart, Kris Van Marcke, Ruurd Pels & Jan Smulders International Insurance Traffic with Software Agents
Session 5C: Agents: Logic & Game Theory • Martin Caminada Agent Dialogues using Hang Yourself Arguments • Mehdi Dastani & Leendert van der Torre What is a Normative Goal? • Jan Broersen, Mehdi Dastani & Leendert van der Torre Logical Specification of Agent Architecture Components • Ann Nowe, Johan Parent & Katja Verbeeck Social Agents Playing a Periodical Policy • Steve Kremer & Jean-François Raskin Game Analysis of Abuse-free Contract Signing
Session 4C: Data Mining II • Jan C. Bioch & Viara Popova Monotone Decision Trees and Noisy Data • Davy Janssens, Tom Brijs, Koen Vanhoof & Geert Wets Evaluating the Performance of Cost-based Discretization versus Entropy- and Errorbased Discretization • Danielle Sent & Linda C. van der Gaag Test Selection: the Gini Index and the Shannon Entropy Behave Differently • Nico Jacobs & Hendrik Blockeel Sequence Prediction with Mixed Order Markov Chains • J.J. Verbeek, N. Vlassis & B. Kröse Coordinating Principal Component Analyzers 12.20 – 13.10 13.10 – 13.40 13.40 – 15.20
15.20 – 15.40 15.40 – 16.40
Session 6A: Game Playing • Erik van der Werf, Jos Uiterwijk & Jaap van den Herik Solving Ponnuki-Go on Small Boards • Pieter Spronck, Ida Sprinkhuizen-Kuyper & Eric Postma Improving Opponent Intelligence by Machine Learning • Mark H.M. Winands, Levente Kocsis, Jos W.H.M. Uiterwijk & H. Jaap van den Herik Learning in Lines of Action
Lunch BNVKI meeting Paper presentations: Session 5
Session 5A: Bayesian Reasoning II • Peter Lucas Restricted Bayesian Network Structure Learning • Michiel van Wezel & Walter Kosters Numerical Integration by Cubature Formulae in Bayesian Neural Networks • Wojciech Zajdel & Ben Kröse Bayesian network for multiple hypothesis tracking • Hilbert J. Kappen & Wim Wiegerinck Novel iteration schemes for the Cluster Variation Method • Peter Antal, Geert Fannes, Yves Moreau & Bart De Moor Using Literature and Data to Annotate and Learn Bayesian Networks.
BNVKI Newsletter
Coffee Break Paper presentations: Session 6
Session 6B: Learning Agents • Kurt Driessens & Saso Dzeroski Integrating Experimentation and Guidance in Relational Reinforcement Learning • Michel van Dartel, Eric Postma & Jaap van den Herik Universal Properties of Adaptive Behaviour • Paul Vogt Anchoring symbols to sensorimotor control
86
August 2002
•
Session 6C: Agents: Applications • Sander Bohte, Enrico Gerding & Han La Poutré Competitive Market-based Allocation of Consumer Attention Space • Catholijn M. Jonker, Jan Treur & Wieke de Vries Temporal Analysis of the Dynamics of Beliefs, Desires, and Intentions • Catholijn M. Jonker, Jacky L. Snoep, Jan Treur, Hans V. Westerhoff & Wouter C.A. Wijngaards Putting Intentions into Cell Biochemistry: An Artificial Intelligence Perspective 16.40 – 17.00
•
You can send your abstract to:
[email protected] or, if e-mail is not possible, to: CLIN 2002 Tanja Gaustad, Alfa-Informatica Rijksuniversiteit Groningen P.O. Box 716 9700 AS Groningen The Netherlands Please take note of the following dates: Deadline for submission: 27 September 2002 Notification of acceptance: 11 October 2002
Awards ceremony & closing (BNAIC Prize for best original paper)
The local organizer of CLIN 2002 is Tanja Gaustad. A volume with proceedings of the twelfth CLIN meeting (held 30 November 2001, in Enschede) will be available at this year's meeting. We intend to produce a volume of the proceedings of CLIN 2002 before CLIN 2003. Papers for these proceedings will have to be written in English; they will be reviewed by a committee to be appointed in due time.
IMPORTANT DATES & LINKS Deadline for early registration: September 15, 2002 For registration information, the conference program and other information, please consult the conference website at: http://www.cs.kuleuven.ac.be/conference/bnaic02/
Call for Papers CLIN 2002
Twelfth Dutch-Belgian Conference on Machine Learning Benelearn 2002
November 29, 2002, Groningen The 13th meeting of Computational Linguistics in the Netherlands will be held in Groningen on the 29th of November, 2002. It will be organized by the Humanities Computing department of the Rijksuniversiteit Groningen in cooperation with SIKS.
December 4, 2002, Utrecht Call for Papers Benelearn is the annual machine learning conference of Belgium and The Netherlands. It serves as a forum for researchers in this field to exchange ideas and present recent work. The official language of the conference is English. Benelearn 2002 will be organized by the Institute of Information and Computing Sciences of the Universiteit Utrecht.
Keynote speaker of this year's meeting will be Hugo Brandt Corstius. The topic of his talk will be announced later. Researchers are invited to present papers on all aspects of computational linguistics (phonetics, phonology, morphology, syntax, semantics, pragmatics, machine translation, computational lexicography, formal languages, grammar formalisms, information retrieval, information extraction, text mining, knowledge representation, parsing and generation, dialogue management, embodied conversational agents, corpus-oriented methods, etc.).
Benelearn wants to broaden its focus to disciplines that are closely related to Machine Learning. We invite papers relevant to machine learning and related disciplines in the broadest sense, especially papers that cross borders between disciplines. Possible paper topics include, but are not limited to: - Applications of Machine Learning - Bayesian Learning - Bayesian Networks - Case-based Learning - Computational Learning Theory - Data Mining
Authors should submit an abstract in English or Dutch (preferably by e-mail, in flat ASCII). The abstract should contain: • a title BNVKI Newsletter
your name, address, affiliation, and e-mail address a short outline of the paper (10-20 lines)
87
August 2002
-
Evolutionary Computation Hybrid Learning Systems Graphical Models Inductive Learning Inductive Logic Programming Knowledge Discovery in Databases Language Learning Learning and Problem Solving Learning by Analogy Learning in Multi-Agent Systems Learning in Dynamic Domains Learning to Search Multistrategy Learning Neural Networks Reinforcement Learning Robot Learning Scientific Discovery
FURTHER INFORMATION Marco Wiering Universiteit Utrecht, IS - Intelligent Systems Group Institute of Information and Computing Sciences Padualaan 14, 3508 TB, Utrecht, The Netherlands
[email protected] http://www.cs.uu.nl/~marco/benelearn2002/
Subsidie Programma Neurale Technologie Beoordelingsprocedure STW-AANVRAGEN binnen de Stichting Neurale Netwerken (SNN) In 2002 zal in samenwerking tussen SNN en STW een subsidieronde op het gebied van neurale netwerken en verwante onderzoeksgebieden worden georganiseerd. Een belangrijke recente ontwikkeling op het gebied van neurale netwerken is de verbreding van het onderzoeksgebied in de richting van de Bayesiaanse statistiek, de probabilistische kunstmatige intelligentie en machine learning. Alle aanvragen dienen het onderzoek aan en/of het gebruik van deze methoden als centraal element te hebben.
Papers are limited to 8 pages (using a 11pt Times Roman font), including figures, title page, references, and appendices. Style files can be downloaded from the conference web site. The papers will be primarily refereed on their relevance to the conference, and according to clarity and overall quality criteria. Email submissions are strongly preferred. Please send a PostScript or PDF file to Marco Wiering (
[email protected]).
De Technologiestichting STW (voorheen Stichting voor de Technische Wetenschappen) is in 1981 opgericht om in de tweede geldstroom het onderzoek te financieren op technischwetenschappelijk gebied. De doelstelling van STW luidt: het geven van steun aan technischwetenschappelijke onderzoeksprojecten van hoge kwaliteit, alsmede het bevorderen van het daadwerkelijk gebruik van de resultaten van dit onderzoek door derden, in casu gebruikers. Sedert 1989 is STW een NWO-stichting met een eigen gebiedsbestuur.
IMPORTANT DATES Deadline for submissions: September 27, 2002 Notification: October 18, 2002 Final version: November 15, 2002 Conference: December 4, 2002 ORGANIZING COMMITTEE Marco Wiering (Universiteit Utrecht) Walter de Back (Universiteit Utrecht)
De doelstelling van SNN is: het stimuleren van fundamenteel en toegepast onderzoek op het gebied van neurale netwerken en verwante onderzoeksgebieden, het verzorgen van kennisoverdracht naar het bedrijfsleven en het exploreren van nieuwe toepassingsgebieden. Bij SNN kunnen, door medewerkers in vaste dienst van Nederlandse universiteiten, projectvoorstellen worden voorgedragen, die op het werkterrein van de SNN liggen en beantwoorden aan de STW-criteria: naast kwalitatief hoogstaand onderzoek een duidelijke technische component en een streven naar een hoge utilisatiegraad (utilisatie: gebruik van de onderzoekresultaten door derden). In deze projectvoorstellen kunnen aanvragen worden gedaan voor aio's, toegevoegd onderzoekers, post-
PROGRAM COMMITTEE Hendrik Blockeel (KU Leuven) Ad Feelders (Universiteit Utrecht) Ben Kröse (UvA) Stephan ten Hagen (UvA) Mannes Poel (TU Twente) Eric Postma (Universiteit van Maastricht) Maarten van Someren (UvA) Dirk Thierens (Universiteit Utrecht) Katja Verbeeck (Vrije Universiteit Brussel) Nikos Vlassis (UvA) Marco Wiering (Universiteit Utrecht) INVITED SPEAKER Helge Ritter (Universität Bielefeld, Germany)
BNVKI Newsletter
88
August 2002
docs, wetenschappelijke medewerkers (programmeurs), materieel en apparatuur.
aantal projectvoorstellen, zullen de voorstellen over twee jury’s worden verdeeld.
Na behandeling in een afzonderlijke ronde binnen SNN worden deze aanvragen ter honorering aangeboden aan het bestuur van de Technologiestichting STW. De behandeling van de aanvragen binnen SNN geschiedt onder verantwoordelijkheid van een door SNN ingestelde commissie. Bij de samenstelling van die commissie wordt rekening gehouden met de thematiek van de ingediende voorstellen. Om die reden is als voorwaarde voor behandeling in de STW-ronde een vooraanmelding van de aanvragen vereist. Aan het secretariaat van SNN dienen voor 1 november 2002 van ieder projectvoorstel schriftelijk enkele kerngegevens te worden doorgegeven. Het betreft hier gegevens over de aanvrager(s), de titel en de probleemstelling van het onderzoek (samen maximaal 1 A4).
Het resultaat van deze werkzaamheden, in de vorm van een honoreringsvoordracht, wordt ingediend bij het bestuur van de Technologiestichting STW. Het honoreringspercentage zal rond de 40% liggen. De planning is erop gericht de honoreringsvoordracht op uiterlijk 1 juni 2003 voor te leggen aan het STW-bestuur. Een eerste uitspraak van het STW-bestuur kan verwacht worden maximaal twee maanden nadat de aanvragen aan het STW-bestuur zijn voorgelegd. Relevante data in de STW-procedure binnen SNN zijn opgenomen in nevenstaand schema. BEOORDELINGSPROCEDURE STW-AANVRAGEN BINNEN SNN 01-11-2002
uiterste datum vooraanmelding bij SNN 01-12-2002 uiterste datum indienen voorstellen bij SNN 01-03-2003 verzending referentencommentaren aan aanvragers 01-04-2003 reacties van aanvragers op referentencommentaren bij STW ontvangen, protocol wordt opgesteld 01-05-2003 voorleggen protocollen aan STW-jury en indiening honorerings-advies bij het STWbestuur Informatie over de Technologiestichting STW kan ingewonnen worden bij het STW secretariaat, zie tevens website http://www.stw/nl
De complete aanvragen moeten 1 december 2002 bij het SNN secretariaat zijn aangeleverd. Deze aanvragen zullen dan naar STW worden doorgestuurd die de verdere procedure administratief begeleidt. De commissie streeft ernaar elke aanvraag voor te leggen aan tenminste vijf anonieme deskundigen, die door de commissie worden uitgekozen op grond van hun kennis van het desbetreffende onderzoeksterrein. Als regel wordt gehanteerd, dat twee deskundigen afkomstig zijn uit de kring van potentiële gebruikers en drie uit die van de universitaire onderzoekers. Aan de referenten wordt gevraagd het projectvoorstel te beoordelen aan de hand van de hieronder beschreven criteria. Er wordt naar gestreefd dat uiterlijk op 15 februari 2003 alle rapporten van referenten binnen zijn op het STW-secretariaat. Vervolgens wordt door het secretariaat per voorstel een aanzet tot protocol van de referentencommentaren opgesteld, waarbij alle opmerkingen per beoordelingscriterium worden geïntegreerd.
CRITERIA VAN TOETSING VAN IN TE DIENEN PROJECTEN Alle aanvragen dienen het onderzoek naar en/of het gebruik van neurale netwerken, Bayesiaanse statistiek, de probabilistische kunstmatige intelligentie of machine learning als centraal element te hebben. Daarnaast worden de projecten beoordeeld op:
Hierna wordt de aanvrager in de gelegenheid gesteld op de kritiek in te gaan en eventuele vragen te beantwoorden. De op schrift gestelde repliek van de aanvrager dient voor 1 april 2003 aanwezig te zijn op het STW-secretariaat. Het secretariaat legt voor ieder project de referentencommentaren en het antwoord van de aanvrager vast in een per criterium gerangschikt protocol. De aanvagen worden vervolgens, tezamen met de protocollen, voorgelegd aan een STW-jury die een rapport-systeem aanlegt, waaruit een honoreringsadvies op basis van prioriteits-volgorde volgt. Gezien het verwachte
BNVKI Newsletter
Wetenschappelijke kwaliteit 1.1 competentie van onderzoekers/onderzoeksteam 1.2 originaliteit van het voorstel 1.3 doeltreffendheid van de onderzoekmethode 1.4 programma 1.5 tijdfasering 1.6 beschikbare infrastructuur 1.7 personeelskosten (aantal, kosten, doelmatigheid) 1.8 overige kosten
89
August 2002
Utilisatie 2.1 praktische toepasbaarheid in industrie, maatschappij, techniek of wetenschap 2.2 mogelijkheden tot ontwikkeling door een (Nederlands) bedrijf van een verkoopbaar product 2.3 belang voor de technische vooruitgang op lange termijn 2.4 concurrentiepositie van de Nederlandse industrie 2.5 octrooipositie van de Nederlandse industrie (indien van toepassing) 2.6 is er een gebruikersgroep? 2.7 zijn er reeds lopende contracten met gebruikers?
Section Editor Richard Starmans SIKS Masterclass on Human Computer Interaction: Design for Cultures on the Move
Zie tevens de STW-brochure "Richtlijnen Open Technologieprogramma". Deze brochure kan schriftelijk of per e-mail worden aangevraagd bij STW of SNN, zie tevens website http://www.stw.nl/stw/richtlijnen/.
August 27, 2002, Amsterdam On August 27, 2002 the School for Information and Knowledge Systems (SIKS) organizes a masterclass with Austin Henderson. Participation (lunch included) is free for all SIKS-members. Location will be the Campus of the Vrije Universiteit Amsterdam.
SECRETARIATEN SNN Universiteit Nijmegen Postbus 9101 6500 HB Nijmegen tel: 024-3614245 fax: 024-3541435 e-mail
[email protected] http://www.snn.kun.nl/
PROVISIONARY PROGRAM -
Technologiestichting STW Postbus 3021 3502 GA Utrecht tel: 030-6001 285 fax: 030-6014 408 e-mail:
[email protected] http://www.stw.nl
-
REGISTRATION If you want to participate, please send an e-mail to
[email protected] and inform us whether you take the lunch or not. Deadline: August 20, 2002
AI EDUCATION
ABSTRACTS
Section Editor Evert van de Vrie
The Design of Socio-Technical Practice The past decades have seen huge improvements in computer systems but these have proved difficult to translate into comparable improvements in the usability and social integration of computers. We believe that the problem is a deeply rooted set of assumptions about how computer systems should be designed, and about who should be doing that design. Human organizations are continually evolving to meet changing circumstances of resource and need. In contrast, computers are quite
M.Sc. Theses in Section AI Education Supervisors of remarkable M.Sc. work are invited to ask their student for a short article, to be submitted to the editor of the Section AI Education.
BNVKI Newsletter
Introduction by Gerrit v.d.Veer (VU) The Design of Socio-Technical Practice by Austin Henderson Changing Requested System Requirements by Johan Hoorn Design for new cultures of use - scenario as a method to bridge the gap by C. Chisalita
90
August 2002
processes, safeguarding the fit between equirements on the one hand and processes and goals on the other. This is done by introducing a mechanism from emotion psychology that explains how events can change attribute values of internal concepts such as objects, tasks, agents, and roles. To bypass the idiosyncrasies of time, an attempt is made to isolate the components of goals and processes that are fixed in so far as they are connected to changing task environments. Such an achievement would be helpful to resolve the mix-up of tongues among the disciplines involved in system design (e.g., ethnography, psychology, arts, ergonomics, and computer science). Identifying the fixed components in changing business processes and goals would add to creating, as it were, a designers' Esperanto.
rigid, incapable of adaptation on their own. Therefore when computer systems are incorporated into human organizations, those organizations must adapt the computers to changing circumstances. This adaptation is another human activity that technology should support, but our design philosophies are oddly silent about it. Central to the problem is the approach that technologists adopt to the design of "systems": by training, interest and practice, they instinctively separate the technical system from its context of use. In particular this very partial viewpoint tends to limit solutions to those that are centrally technical, and is blind to users as system elements that are fully connected to the changing needs and values of the circumstances that the system serves. This talk explores the origins of these problems in the norms developed for managing human organizations, proposes partial solutions that can be implemented with current systems technology, and speculates about the longterm potential for radical improvements in system design. It proposes that the evolution of sociotechnical practice must be fundamentally "pliant".
About the speaker: Johan Hoorn has a background in Psycho-linguistics. He is currently a Postdoc in the Software Engineering Department of the Vrije Universiteit, working in a project to relate business goals and business processes to requirements engineering and task analysis.
About the speaker: Austin Henderson's background is in computer science (Ph.D., MIT), with experience in design and ethnography. His 30-year career in Human-Computer Interaction includes user interface research and architecture at Bolt Beranek and Newman, Xerox research (both PARC and EuroPARC) and Apple Computer, and strategic industrial design with Fitch. He has worked in small research terms, managed research projects and laboratories, and led larger multi-divisional collaborations. Currently he is the principal of Rivendel Consulting & Design, working with corporate customers on interface, product and strategic design, with a focus on explorations of illformed and emerging questions through observation, interactional interviewing, and reflective intervention. These methods develop rich access to participants' and clients' views of their activity. This serves as a basis for co-producing new perspectives and technical frameworks, and addressing the issues in the design of sociotechnical systems and practice. Austin is also a principal in Pliant Research, a research effort pursuing basic research in overcoming formality in computing. Professionally, Austin has participated in the activities and leadership of ACM's Special Interest Group on Human-Computer Interaction (ACM/SIGCHI) for over 15 years, including serving as Chair.
Design for new cultures of use - scenario as a method to bridge the gap In design for novel situations the main issue frequently turns out to be the step from the current situation to the intended context of use. If the artifact to be developed would ever be implemented, this would change the world of the prospective users. Alternatively, or the design focuses on future situations, where the context of use will be different of today's world anyhow. The main message of our approach is that design in such cases should step beyond the specification of the artifact. In fact the base of the specification should be the envisioning of the context of use. In the case of interactive systems this frequently turns out to be a radical change from the existing situation. New technology in many cases effects the human organization, as new functionality results in restructuring human roles. Moreover, technology effects procedures of work, and often comes with changes in the physical conditions of work (think of mobile telephones). We propose to look at this situation as design for new cultures of use. In fact, the context of use in these cases is equivalent with a new work culture. Only if we have ideas of what the new situation of use could be, and what this would mean for the intended user, would it make sense to focus on any details of technology. Meaning in fact is created differently in different cultures of use based on the common ground in the context of use, so we need to make sure that we take this into account if we investigate the understanding of
Changing Requested System Requirements To facilitate an adequate requirements management, concepts of Groupware Task Analysis are connected to fluctuating business goals and
BNVKI Newsletter
91
August 2002
setting and to inform them about current developments and some new activities and plans for the coming year. This year a small scientific symposium will be organized at the SIKS day as well. Three guest speakers have agreed to participate:
prospective users. At least the intended context of use should be made clear to these users. So we need to confront prospective users, or people who can be considered equivalent to these, with a vision of this new context and culture of use, even before any details of the technology have been developed. Designing for novel cultures of use requires the investigation of the meaning of the artifact in use for the prospective users.
Prof.dr. A. Nijholt (UT) Prof.dr. H. Akkermans (VU) Prof.dr. P. Doherty (Linköping University) Titles and abstracts of the talks will be made available soon. All members of our research school (research fellows, associated members and Ph.D. students) as well as the members of SIKS' Advisory Board are invited to join the SIKS day 2002.
Scenarios trigger people to develop mental models of imaginary worlds, and are, hence, ideal vehicles to study users' mental models of technology in use long before the details of the technology itself are specified. This approach focuses on the future world as understood by the user, not on the feasibility of the technical solutions.
Participation (lunch included) is free, registration is required.
We will go in to detail about the requirements of scenarios for the type of early design we discussed. We will refer to several types of scenarios as elaborated in the literature (e.g., Carroll's recent books in this field). The scenarios as we need for this design process are generally vague about aspects of the technology (where functionality is the only part that matters in the early cycles of design). We will show examples where the scenario was prepared by sketching the context of use and providing a mock up of a device (e.g., a cardboard box with some unlabelled buttons) where the prospective users were invited to act out their interpretation of actual use.
Registration: A registration form can be found at www.siks.nl
SIKS Masterclass: Security and Privacy in Cyberspace November 4, 2002, Amsterdam This SIKS Masterclass takes place at the Vrije Universiteit Amsterdam. PROVISIONARY PROGRAM
Finally, we will show some examples of this approach as applied in design in industry and in public administration design, where we have the possibility to participate in design activities that develop in parallel (and mostly are related to) changes in the organizational culture.
10:00-11:00:
About the speaker: Cristina is a Ph.D. student at the Department of Software Engineering of the Vrije Universiteit, Amsterdam. She graduated in Cognitive and Organizational Psychology at the University of Cluj, Romania.
11:00-12:00:
SIKS Day 2002
14:00-15:00:
November 1, 2002, Utrecht On November 1, the School for Information and Knowledge Systems organizes its annual SIKS day. The location will be the City castle Oudaen in Utrecht (http://www.oudaen.nl/). The main aim of the event is to give SIKS members-participating in research groups all over the country- the opportunity to meet each other in an informal BNVKI Newsletter
92
Prof. Reind van de Riet (host, VU): Introduction on S&P in Cyberspace; S&P and Workflow alter egos for keeping privacy in Cyberspace Prof. Martin Olivier (RAND University Johannesburg, South Africa): security in federated databases - schema integration - addition of a federated 'layer' - policy mediators/'contract' enforcers, etc Prof Ehud Gudes (Ben Gurion University, Ber Sheba, Israel): Security techniques in the form of crypto systems: - Encryption, symmetric vs. public keys, RSA. - Digital signatures, Hash functions - The notion of certificates
August 2002
-
15:00-16:00:
Authentication protocols, including overview of the Kerberos protocol Radu Serban (VU) on Fireballs and FBML; Overview of his Dissertation titled: The Private Cyberspace: Modelling electronic environments inhabited by privacyconcerned agents
IMPORTANT DATES Deadline for abstracts: September 16, 2002 Notification of acceptance: October 7, 2002 Deadline for final copies: November 7, 2002 Workshop: December 6, 2002 REGISTRATION Participants are kindly asked to register online on the workshop home page. Attendance to the workshop is free. For participation at the lunch, a fee of EUR 20 will be asked. For SIKS Ph.D. students participation at the lunch is free as well. They should indicate that they are a member on the electronic form that can be found at: http://www.law.kuleuven.ac.be/icrinew/subscription .php?id=7. In exchange for publicity on the workshops announcements and website, company representatives who attend the workshop are invited to donate a contribution of EUR 100.
DIR-2002 Third Dutch-Belgian Information Retrieval Workshop December 6, 2002, Leuven The third Dutch-Belgian Workshop on Information Retrieval will take place on Friday 6/12/2002 at the Katholieke Universiteit Leuven (Belgium). The keynote speaker will be Prof. Dr. Karen SparckJones (Computer Laboratory, Cambridge University). The primary aim of the workshop is to provide a meeting place where researchers from the Low Countries (and neighbouring countries) who are working in the domain of information retrieval and related disciplines can exchange information and present new research developments.
PROGRAM CMMITTEE Marie-Francine Moens (KU Leuven), Djoerd Hiemstra (TU Twente), Wessel Kraaij (TNO Delft) This workshop is organized by the Interdisciplinary Centre for Law and IT in cooperation with IWT Vlaanderen and Research school SIKS. More information: http://www.law.kuleuven.ac.be/dir2002/
CALL FOR ABSTRACTS Authors who are interested in presenting original research at the workshop are invited to submit an extended abstract. Abstracts should be maximum 2 pages long and should be submitted September 16, 2002. Depending on their number, the submissions will be reviewed by two or three domain experts.
SECTION KNOWLEDGE SYSTEMS IN LAW AND COMPUTER SCIENCE
Abstracts can cover any topic related to IR: - Retrieval models, language models - Text representation, information extraction, text summarization, text categorization, topic tracking and event detection - Query processing, thesaurus construction, user models - Cross-language IR - IR for the Web, XML and metadata, link analysis - Multimedia IR, video retrieval, audio retrieval - Question answering systems - Domain-specific IR applications - IR evaluation Abstracts should be sent to
[email protected] as a postscript or PDF file.
BNVKI Newsletter
Section Editor Marie-Francine Moens Feasibility of Reliable (Semi)-Automatic Anonymizing of Court Decisions Presentation of Jacqueline Dake JURIX meeting, May 29, 2002 Report by Marie-Francine Moens ICRI, KU Leuven Court decisions are very valuable for legal research and are especially useful for reasoning with precedents. However, the decisions can only be anonymously stored in databases that are publicly 93
August 2002
uncertainty about the class label of an example. The entropy is zero when all examples of a set belong to the same class in which case there is no uncertainty. The more classes there are and the more the classes are equally distributed the larger the entropy becomes. The information gain of a feature measures the expected reduction in entropy by partitioning the examples according to this feature. The information gain thus measures the quantity of information that the feature contributes to the knowledge of the correct class label. In the experiments the gain ratio is used, which is the normalization of the information gain for the number of values of a feature and which reduces an overestimation of the information gain when a feature has many values. When searching for an analogy between two situations, it is important to determine which aspects of the two situations are relevant to form an analogy between them. The weighting of the features accomplishes this effect. To determine the distance between two instances, it is assumemed that a numerical similarity function is used that combines the importance or weight for each feature with its degree of match and sums the scores to create a match score for each comparison.
accessible. Anonymizing the decisions is currently manually done, which is an important bottleneck for the construction of the databases. Jacqueline Dake gave a very engaging lecture about her research. The research regards the identification (i.e., the classification) with the technique of memory-based learning of phrases and words in court decisions that need to be rendered anonymous. The research was part of a project during the study of Taal en Kunstmatige Intelligentie at the Katholieke Universiteit Brabant (KUB). Paul Huygen of the Vrije Universiteit Amsterdam and Hans Paijmans of the KUB supervised the research. BASICS Memory-based learning (also called instance-based learning or k-nearest neighbor classification) learns from classified examples. This technique will not generalize the examples into an explicit abstract function or rule that is applied on the new instances to be classified, but rather is a “lazy” machine learning technique: It only stores the classified examples. When a new instance is classified, it is compared with all the classified examples stored and the distance with each of these examples is computed. The new instance is granted the classes of the k closest classified examples. This is a form of learning by analogy, which when learning patterns in natural language, often yields good results because natural language has a lot of exceptions besides patterns that are easily recognizable.
RESULTS The classifier was trained and tested upon 93 civil cases of the court of 's Gravenhage. This small corpus contains 1208 words referring to private data upon a total of 164911 words. Tenfold crossvalidation of this data set and the use of the memory-based learner with k = 1 gives the following values of recall, precision and F-measure.
FINDING PRIVATE DATA ANON NOTAG
In the experiments of Jacqueline Dake, each word of a decision is classified as ANON (anonymous) or as NOTAG. Each word in the decisions of the training and the test set is represented as a feature vector. The features or attributes can represent the properties of the word itself (e.g., whether or not the word starts with a capital letter, whether it contains a hyphen, whether it is composed of numerical characters) or represent the features of the words that precede or follow it (e.g., whether the phrase that precedes is “in zake” which might indicate that the name of the plaintiff follows). For the experiments, Jacqueline Dake had only access to court decisions in which the private data were already removed. This means that only features with regard to the context of the word that is or is to be classified are included in the feature vectors.
Precision 60.35% 99.66%
F-measure 57.03% 99.70%
Recall is the percentage of actual class members that are recognized. Precision is the proportion of members assigned to the class that really are class members. The F-measure combines both recall and precision. Accuracy measures the number of correct assignments upon the total number of assignments. The global accuracy is 99.40%. Jacqueline Dake is pleased with the results considering the limitation of available classification features and encourages further research. For instance, she proposes to include part-of-speech (POS) and noun phrase chunk information in the feature vectors. The purpose of the research was to build a reliable classifier and to learn the degree of reliability. Two minimum distances are defined, one for reliable class assignment and one for a questionable assignment. The latter can be used in combination
The features are weighted with techniques from information theory. The entropy expresses the amount of uncertainty, for instance, the amount of BNVKI Newsletter
Recall 54.06% 99.74%
94
August 2002
with a manual control and correction (cf. semiautomatic anonymizing).
METADATA VOOR WETGEVING Er zijn acht zogenaamde metadata die essentieel zijn voor de elektronische verwerking en opslag van wetgeving, zoals het aktenummer, het artikelnummer, het territoriaal toepassingsgebied (het Nederlands, Frans, Duits taalgebied en het tweetalig gebied Brussel hoofdstad, het laatste zowel voor de Nederlands- als voor de Franstaligen) en het versienummer. Met betrekking tot de metadata onderscheidt Keymis drie verschillende situaties. In de eerste plaats kunnen de metadata overeenstemmen met aanwezige, duidelijk identificeerbare referentiegegevens in de huidige wetgevende akten (bijv. artikelnummer). In de tweede plaats zijn er metadata die overeenstemmen met aanwezige, niet duidelijk te identificeren referentiegegevens in de huidige wetgevende akten (bijv. territoriaal toepassingsgebied). In de derde plaats zijn er metadata zonder overeenstemmende referentiegegevens in de huidige wetgevende akten (bijv. versienummer). Vooral voor de tweede situatie moet men alert zijn. Elektronische verwerking blijft problematisch zonder adequate verduidelijking van de referentiegegevens tot bruikbare essentiële metadata.
The technique is applicable for rendering private data in court decisions anonymous. In addition, it is useful for any information extraction task that semantically classifies phrases and words in a text. However, the limited number of available examples that are manually annotated and the confrontation with the rich variety in expression possibilities in natural language are annoying limitations. OTHER SOLUTIONS? The lecture was followed by a vivid discussion. The question was asked whether the drafters of the decisions can mark or tag the private data (e.g., the names of the plaintiff and defendant) or type these data in designated fields of template documents. It was remarked that a large number of the private data occur in the free text of the decision (e.g., in the motivation part). This does not guarantee that the drafter of the text would consistently mark these data. In addition, up until today judges and court clerks resist using structured templates when drafting decisions. Consequently, intelligent techniques will continue to exist for the identification and processing of data in court decisions.
WORKFLOW VAN EEN WETSONTWERP Een WFM-systeem kan worden ingezet om de op dit moment papieren gegevensstroom bij het ontwerpen van wetgeving om te zetten in een elektronische. Als voorbeeld van het wetgevingsproces wordt geschetst hoe dit voor een wetsontwerp op federaal niveau gebeurt. Het proces vangt aan met de voorbereiding van de wetgeving door externen of door ambtenaren van het betreffende wetgevingsdepartement. Voordat de wet een officieel wetsvoorstel is dat kan worden ingediend in de Kamer gaat het langs de ministerraad en voor advies naar de Raad van State, maar in uitzonderlijke gevallen kunnen beide gepasseerd worden. In een WFM-systeem zal uiteraard ook met deze uitzonderlijke situaties rekening moeten worden gehouden. De daadwerkelijke workflow is gecompliceerder, zo vindt er bijvoorbeeld in bepaalde gevallen terugkoppeling naar de ministerraad plaats nadat de Raad van State om advies is gevraagd. Nadat het wetsontwerp is ingediend gaat het wetsontwerp in de Kamer doorgaans langs een commissie met expertise op het betreffende terrein alvorens plenair behandeld te worden. Ook tijdens deze procedure bij de Kamer kan nog advies worden gevraagd bij de Raad van State. Uiteindelijk komt de wet, na eventuele amendering en behandeling in de Senaat voorzien van de nodige handtekeningen van minister(s) en Koning, weer terug bij hetzelfde
Wetgevende Procedures en de Kwaliteit van Wetgeving in een Automatiseringscontext Presentatie van Dirk Keymis JURIX-bijeenkomst op 29 mei 2002 Verslag door Arno R. Lodder CLI, Vrije Universiteit Dirk Keymis is lid van LIIR, de onderzoeksgroep Legal Informatics and Information Retrieval dat deel uitmaakt van het Interdisciplinair Centrum voor Recht en Informatica (ICRI) van de KU Leuven. De vraag die vandaag centraal staat is of voor de toenemende hoeveelheid wetgeving automatisering een oplossing is. In de presentatie wordt een analyse gegeven van het huidige wetgevende proces en wordt aangegeven hoe een workflowmanagent-systeem (WFM-systeem) hierbij zou kunnen worden ingezet. Tevens wordt ingegaan op de juridische kwaliteitseisen waar de automatisering die in het wetgevende proces gebruikt wordt aan zou moeten voldoen.
BNVKI Newsletter
95
August 2002
aan elektronische ontsluiting en waarvoor nieuwe kwaliteitseisen ontwikkeld moeten worden. Hierbij moet gedacht worden aan het consequent en consistent toekennen van metadata. Keymis stelt verder voor om een centrale, coördinerende instantie op te richten. Deze instantie moet kwaliteitseisen formuleren, toezien op de naleving daarvan en toezien op een finale controle van de wetgevende akte die de aanwezigheid van de benodigde essentiële metadata voor verdere verwerking garandeert. Bij de finale controle moet geëxperimenteerd worden met de nodige technologie om de elektronische verwerking en controle automatisch te laten verlopen. Wanneer een keuze gemaakt is kan de technologie door de centrale instantie via de wetgevende procedures verspreid worden.
wetgevingsdepartement dat instaat voor de finalisering en het doorsturen van de akte naar het Belgisch Staatsblad. Binnen het proces van wetgeving zijn er vijf categorieën controles die plaatsvinden: wetstechnische controle, juridische controle, procedurele contole, administratieve controle en taalkundige controle. Alle controles zijn van belang voor de goede werking van het huidige systeem. De essentiële metadata die overeenstemmen met beide soorten van aanwezige referentiegegevens worden direct of indirect in de eerste drie categorieën van controles geïdentificeerd. Het WFM-systeem moet zich dan ook in eerste instantie richten tot deze drie categorieën. De controles die op dit moment plaatsvinden zijn niet toereikend om elektronische opslag en verwerking mogelijk te maken. Problemen zijn onder andere gebrek aan geconsolideerde wetteksten. Er is ook een duidelijk gebrek aan eenvormige elektronische versies van documenten.
CONFERENCES, SYMPOSIA WORKSHOPS
CENTRAAL REGELEN
Below, the reader finds a list of conferences and web sites or email addresses for further information.
Om de problemen te lijf te gaan (en dus te komen tot een goed WFM-systeem) stelt Keymis voor aansluiting te zoeken bij de expertise van bestaande initiatieven inzake wetgevingskwaliteit, maar merkt op dat er problemen zijn die uitsluitend verwant zijn
BNVKI Newsletter
SEPTEMBER 3-6, 2002 2002 IEEE Symposia on Human Centric Computing Languages and Environments (HCC’02). Arlington, USA. http://www2.cs.fau.de/HCC02/cfp.html 96
August 2002
SEPTEMBER 5-6, 2002 ICT-kenniscongres. Den Haag, The Netherlands. http://www.ict-kenniscongres.nl
JANUARY 21-22, 2003 International Workshop on Technology. Nagasaki, Japan. email:
[email protected]
SEPTEMBER 9-13, 2002 IEEE Joint Conference on Requirements Engineering (ICRE’02). University of Essen, Germany. http://www.re02.org
Image
FEBRUARY 4-8, 2003 LINZ 2003 - 24th Linz Seminar on Fuzzy Set Theory. Linz, Austria. http://www.flll.jku.at/linzSeminars/linz2003/ MARCH 24-26, 2003 AAAI Spring Symposium on Agent-mediated Knowledge Management. Stanford University, USA. Information: http://www.dfki.uni-kl.de/~elst/AMKM/index.html
SEPTEMBER 23-26, 2002 JELIA'02: 8th European Conference on Logics in Artificial Intelligence. Cosenza, Italy. http://www.unical.it/jelia/ OCTOBER 1-4, 2002 13th International Conference on Knowledge Engineering and Knowledge Management EKAW02. Sigüenza, Spain. http://babage.dia.fi.upm.es/ekaw02/ekaw02.htm
JUNE 23-27, 2003 Ninth International Conference on Artificial Intelligence and Law (ICAIL 2003). Edinburgh, Scotland, UK. http://www.cirfid.unibo.it/~agsw/ICAILS03
OCTOBER 21-22, 2002 14th Belgian-Dutch Conference on Artificial Intelligence (BNAIC 2002). Leuven, Belgium. http://www.cs.kuleuven.ac.be/conference/bnaic02
ADVERTISEMENT POLICY OF THE BNVKI
OCTOBER 25, 2002 1st EUNITE workshop on Intelligent systems in medical diagnosis and therapy. Coventry University, Coventry, UK. http://www.ctac.mis.coventry.ac.uk/eunite2002.html
1. We allow announcements and advertisements to be inclued into the Newsletter at the cost of € 275,- per full page and € 180,- per half page of one issue. The subscriber receives three proof copies. 2. We allow announcements and advertisements to be included into the Newsletter at the cost of € 450,- per full page and € 320,- per half page for two subsequent issues. The subscriber receives three proof copies of each issue. 3. We allow announcements and advertisements to be included into the Newsletter at the cost of € 900,- per full page and € 680,- per half page for six issues (one Volume). The subscriber receives three proof copies of each issue.
DECEMBER 9-13, 2002 Artificial Life VIII. UNSW, Sydney, Australia. http://alife8.alife.org/ DECEMBER 4, 2002 Benelearn 2002 - Twelfth Dutch-Belgian Conference on Machine Learning. Utrecht, The Netherlands. http://www.cs.uu.nl/~marco/benelearn2002/ DECEMBER 10-12, 2002 The Twenty-second SGAI International Conference on Knowledge Based Systems and Applied Artificial Intelligence (ES2002). Cambridge, UK. http://www.bcs-sges.org/es2002/
The contents of the text offered must be related to AI. The Editorial Board has the right to change the text or to refuse it, if they believe the publication intended does not fit the Newsletter. Moreover, the Board of the BNVKI has decided upon the following alternatives. 4. We allow promotional leaflets to be included into the membership mailing of the Newsletter. The costs for mailing are € 365,-
DECEMBER 11, 2002 First Annual BCS Prize for Progress Towards Machine Intelligence. Cambridge, UK. http://www.bcs-sges.org/es2002/miprize/ JANUARY 12-15, 2003 IUI 2003 The Seventh International Conference on Intelligent User Interfaces. The Palms South Beach Hotel, Miami Beach, Florida, USA. http://www.iuiconf.org/
BNVKI Newsletter
Advanced
97
August 2002
AIO-PLAATS: REDENEREN MET MEDISCHE INTENTIES Ten behoeve van het Medical Information Agent (MIA) project, dat uitgevoerd wordt door de capaciteitsgroepen Medische Informatica en IKAT van de Universiteit Maastricht en door het Centrum voor Wiskunde en Informatica, is er bij de capgroep Medische Informatica een plaats voor een AIO, die zich zal bezighouden met (a) onderzoek naar methoden voor de representatie van verschillende typen intenties van richtlijnen (voor het handelen van medici en verpleegkundigen), (b) het vastleggen van de intenties behorend bij richtlijnen voor de behandeling van oncologische patienten in een besluitvormingsondersteunend systeem, gekoppeld aan een electronisch patientendossier en (c) het valideren van de voorgestelde methoden in de praktijk. Gevraagd wordt een pas afgestudeerde (medisch) informaticus of iemand met een vergelijkbare opleiding met belangstelling voor kennisrepresentatie. Bekendheid met objectgeoriënteerd programmeren is een vereiste. Het project maakt deel uit van het door NWO gesubsidieerde ToKeN2000 programma (www.token2000.nl). Inlichtingen: A. Hasman (
[email protected]) of F. Wiesman (
[email protected]) Sollicitaties: Sollicitaties met CV kunnen gestuurd worden naar: prof.dr.ir. A. Hasman, Universiteit Maastricht, Capgroep Medische Informatica, Postbus 616, 6200 MD Maastricht. Verdere informatie Binnen het MIA project wordt een besluitvormingsondersteunend systeem ontworpen dat het handelen van de arts/verpleegkundige, zoals dit is vastgelegd in het EPD, indien dit handelen niet in overeenstemming is met de er voor geldende richtlijnen, voorziet van commentaar. Op deze manier bevordert het systeem de kwaliteit van het handelen van arts of verpleegkundige. Richtlijnen kunnen vaak op verschillende manieren, afhankelijk van de praktijkomstandigheden, worden uitgevoerd. Het systeem moet kunnen herkennen welke handelwijzen naar de geest van de richtlijnen zijn en de arts/verpleegkundige pas commentaar leveren, als het handelen niet in de geest van de richtlijnen is. In de huidige systemen zijn alleen de richtlijnen geïmplementeerd en kan het dus voorkomen dat arts of verpleegkundige commentaar van het systeem krijgen, dat ze niet (letterlijk) volgens de richtlijnen werken, ondanks het feit, dat de arts/verpleegkundige naar de geest van de richtlijnen handelt. Als het besluitvormingsondersteunend systeem te vaak dit soort meldingen geeft zal het systeem niet meer worden gebruikt. In het MIA project is een van de onderzoeksdoelen het vinden van een geschikte representatie voor de intenties van richtlijnen. Daartoe moet worden nagegaan welke typen intenties er zijn en welke wijze van representatie het beste is voor elk van die typen. Ook zullen de verschillende handelingen moeten worden gerepresenteerd op een dusdanige wijze, dat een systeem kan beredeneren dat een handeling volgens de intenties van de richtlijn is uitgevoerd. Als de intenties en handelingen via een nog te bepalen representatie vastgelegd kunnen worden, dan kan een besluitvormingsondersteunend systeem worden ontwikkeld dat, op basis van de intenties van de richtlijnen, zoals deze in het kennisbestand zijn vastgelegd en de handelingen, zoals deze zijn vastgelegd in het EPD, kan nagaan of de arts naar de geest van de richtlijnen werkt. Zo niet, dan zal het systeem de arts/verpleegkundige voorzien van feedback. Reeds ontwikkeld is een kennisacquisitiemodule, waarmee richtlijnen kunnen worden ingevoerd. Deze kennisacquisitiemodule moet worden uitgebreid met een module, waarmee intenties kunnen worden ingevoerd. De uitvoer van de kennismodule kan direct worden vertaald in een run-time besluitvormingsondersteunend systeem. De wijze waarop dit gebeurt, zal ook aangepast moeten worden om rekening te kunnen houden met de intenties. Er is op dit gebied samenwerking met de Universiteit Eindhoven en met een groep van de Stanford University in de VS. BNVKI Newsletter
98
August 2002
Dr. J. van Looveren (editor Belgium) Vrije Universiteit Brussel, AI Lab Pleinlaan 2, 1050 Brussel, Belgium Tel.: + 32 6293702. E-mail:
[email protected]
ADDRESSES BOARD MEMBERS BNVKI Prof.dr. J.N. Kok (chair) Wiskunde en Natuurwetenschappen, Dept. of Computer Science Universiteit Leiden, Niels Bohrweg 1, 2333 CA Leiden Tel: + 31 71 5277057. E-mail:
[email protected]
Dr. R.J.C.M. Starmans (section editor) Manager Research school SIKS, P.O. Box 80089 3508 TB Utrecht Tel.: + 31 30 2534083/1454. E-mail:
[email protected]
Dr. R. Verbrugge (secretary) Rijksuniversiteit Groningen, Cognitive Science and Engineering Grote Kruisstraat 2/1, 9712 TS Groningen. Tel.: + 31 50 3636334. E-mail:
[email protected]
Ir. E.M. van de Vrie (section editor) Open Universiteit Nederland, Opleiding Informatica Postbus 2960 6401 DL Heerlen Tel: + 31 45 5762366. Email:
[email protected]
Dr. W. van der Hoek (treasurer) Universiteit Utrecht, Department of Computer Science P.O. Box 80089, 3508 TB Utrecht Tel.: + 31 30 2533599. E-mail:
[email protected]
HOW TO SUBSCRIBE
Dr. A. van den Bosch Katholieke Universiteit Brabant, Taal- en Literatuurwetenschap Postbus 90153, 5000 LE Tilburg Tel.: + 31 13 4668260. E-mail:
[email protected]
The BNVKI/AIABN Newsletter is a direct benefit of membership of the BNVKI/AIABN. Membership dues are € 34,03 for regular members; € 22,69 for doctoral students (AIO's); and € 17,88 for students. In addition members will receive access to the electronic version of the European journal AI Communications. The Newsletter appears bimonthly and contains information about conferences, research projects, job opportunities, funding opportunities, etc., provided enough information is supplied. Therefore, all members are encouraged to send news and items they consider worthwhile to the editorial office of the BNVKI/AIABN Newsletter. Subscription is done by payment of the membership due to RABO-Bank no. 11.66.34.200 or Postbank no. 3102697 for the Netherlands, or KBC Bank Veldwezelt No. 457-6423559-31, 2e Carabinierslaan 104, Veldwezelt, Belgium. In both cases, specify BNVKI/AIABN in Maastricht as the recipient, and please do not forget to mention your name and address. Sending of the BNVKI/AIABN Newsletter will only commence after your payment has been received. If you wish to conclude your membership, please send a written notification to the editorial office before December 1, 2002.
Prof. M. Denecker Katholieke Universiteit Leuven Dept. of Computer Science, Celestijnenlaan 200A 3001 Heverlee, België Tel.: + 32 16327544. E-mail:
[email protected] Dr. C. Jonker Vrije Universiteit Amsterdam, Dept. of Artificial Intelligence De Boelelaan 1081, 1081 HV Amsterdam Tel.: + 31 20 4447743. E-mail:
[email protected] Prof.dr.ir. J.A. La Poutré Centrum voor Wiskunde en Informatica P.O. Box 94079 1090 GB Amsterdam Tel.: + 31 20 592 9333. E-mail:
[email protected] Dr. F. Wiesman Universiteit Maastricht, IKAT Postbus 616, 6200 MD Maastricht Tel.: + 31 43 3883379. E-mail:
[email protected]
COPY The editorial board welcomes product announcements, book reviews, product reviews, overviews of AI education, AI research in business, and interviews. Contributions stating controversial opinions or otherwise stimulating discussions are highly encouraged. Please send your submission by E-mail (MS Word or text) to
[email protected].
Drs. B. Zinsmeister Cap Gemini Ernst & Young/Service Practice Bolesian Postbus 2575 3500 GN Utrecht Tel.: + 31 30 6893394. E-mail:
[email protected]
ADVERTISING It is possible to have your advertisement included in the BNVKI/AIABN Newsletter. For further information about pricing etc., see elsewhere in the Newsletter or contact the editorial office.
EDITORS BNVKI NEWSLETTER Dr. F. Wiesman (editor in chief) (See addresses Board Members)
CHANGE OF ADDRESS
Dr. E.O. Postma Universiteit Maastricht, IKAT Postbus 616, 6200 MD Maastricht Tel: + 31 43 3883493. E-mail:
[email protected]
The BNVKI/AIABN Newsletter is sent from Maastricht. The BNVKI/AIABN board has decided that the BNVKI/AIABN membership administration takes place at the editorial office of the Newsletter. Therefore, please send address changes to:
Prof. dr. H.J. van den Herik Universiteit Maastricht, IKAT Postbus 616, 6200 MD Maastricht Tel.: + 31 43 3883485. E-mail:
[email protected]
Editorial Office BNVKI/AIABN Newsletter Universiteit Maastricht, Hazel den Hoed, Dept. Computer Science, P.O. Box 616, 6200 MD Maastricht, The Netherlands E-mail:
[email protected] http://www.cs.unimaas.nl/~bnvki/
Dr. Marie-Francine Moens (section editor) KU Leuven, Interdisciplinair Centrum voor Recht & Informatica Tiensestraat 41 3000 Leuven, België Tel.: + 32 16 325383 E-mail:
[email protected]
BNVKI Newsletter
99
August 2002
Bolesian Cap Gemini Ernst & Young heeft een grote groep specialisten op het gebied van kennistechnologie, kennismanagement en andere interessante onderwerpen op ICT-gebied. Zij werken in de kennistechnologie-practice Bolesian en houden zich voornamelijk bezig met innovatieve projecten. Hierbij kan gedacht worden aan: intelligente matching, bijvoorbeeld resource- en reisplanning, vraagbaaksystemen, beoordelingssystemen binnen financiële instellingen en intelligente planning en schedulingsystemen binnen industrie. Groeien op een goede voedingsbodem Bewust omgaan met kennis en kansen, dat is het kenmerkend cultuurelement en een goede voedingsbodem voor uitdagende werkzaamheden bij de practice Bolesian. Cap Gemini Ernst & Young stimuleert haar medewerkers om hun kennis verder uit te bouwen en daarna uit te dragen binnen bijvoorbeeld interne werkgroepen, de zogenaamde Special Interest Groups. Uitwisseling van kennis met universiteiten en hogescholen behoort tot onze dagelijkse praktijk. Medewerkers van Bolesian participeren actief in onderzoek en onderwijs. Uitdragen van kennis trekt nieuwe kennis aan en brengt onze organisatie in contact met toonaangevende marktpartijen. Niet voor niets verwachten we van onze specialisten dat zij als spreker aanwezig zijn op congressen en dat zij publiceren in de vakbladen. Iedereen bij Cap Gemini Ernst & Young, in welke functie dan ook, kan suggesties doen en initiatieven ontplooien die een verrijking vormen voor Cap Gemini Ernst & Young en haar opdrachtgevers. Verantwoordelijkheid dragen en doorgroeien Werken bij Cap Gemini Ernst & Young moet blijven boeien. Daarom zorgen we ervoor, dat binnen relatief kleinschalige projecten met een platte organisatiestructuur, professionals in een goede sfeer met elkaar samenwerken. Daarbij wordt van elke medewerker verwacht dat hij of zij verantwoordelijkheid draagt. Op die manier wordt doorgroeien een vanzelfsprekendheid. Maar ook buiten het eigen project worden medewerkers steeds op de hoogte gehouden. Daarvoor organiseren we bedrijfs- en projectpresentaties. De Sociale Commissie organiseert vier maal per jaar avontuurlijke activiteiten, waarbij collega’s elkaar ook op een andere manier leren kennen. Zo zie je onze kennistechnologen ook eens waterskiën, karten of op survival. CAP GEMINI ERNST & YOUNG Cap Gemini Ernst & Young behoort wereldwijd tot de grootste managementadvies- en IT-servicesorganisaties en neemt met deze gecombineerde expertise een unieke positie in. De onderneming is in meer dan dertig landen actief op het gebied van business- en IT-strategie, managementconsulting, technologieontwikkeling en outsourcing. Deze diensten worden geïntegreerd aangeboden vanuit marktgerichte sectoren waarin branchekennis is gebundeld. Mede dankzij het multinationale karakter van de onderneming beschikt Cap Gemini Ernst & Young over een bijna onuitputtelijke bron van kennis, methodieken en innovatieve oplossingen. Door hergebruik hiervan en door intensief kennismanagement wordt de time-to-market voor klanten verkort en kunnen zij hun marktvoorsprong vergroten. Cap Gemini Ernst & Young telt wereldwijd meer dan 60.000 medewerkers van wie 7.000 in Nederland en behaalde in 2000 een omzet van 8,5 miljard euro. Meer informatie Meer informatie over specifieke dienstverlening, kantoren en research is beschikbaar op: www.cgey.com en www.nl.cgey.com of stuur een mailtje naar
[email protected].