The 2007 IEEE / WIC / ACM International Conference on Web Intelligence
WI 2007
Home WI 2007 Contact Sitemap IAT 2007 WI2007
What's New
About WI 2007
Important Dates
Submissions
Topics
Call for Papers
Tutorial
Workshop
Workshop
Invited Speakers
Final Program
Organization
Program Committee
Program C
ommittee
Program C
ommittee
Conference Site
About Silicon Valley
Sponsors and Benefits
Related Links
Downloads
Tutorials
Co-located with IAT 2007, GrC 2007, BIBM 2007 conferences

Tutorials

Industrial session on mashups
Organizer : Howard Ho, IBM USA
Presenters: IBM Almaden Research Center, Yahoo, Google and Microsoft

Web-based Support Systems (Tutorial notes)
Dr. JingTao Yao
University of Regina, Canada

Web and Text Mining for Opinion/Trend Analysis (Tutorial notes)
Dr. Lipika Dey
Tata Consultancy Services, India

Agent-Mining Interaction and Integration (Tutorial notes)
Dr Longbing Cao
University of Technology, Sydney, Australia

Trust Mechanisms for Agent Systems (Tutorial notes)
Dr. Sandip Sen
University of Tulsa, USA

Peer-to-Peer Distributed Data Mining for Multi-Agent Applications (Tutorial notes)
Dr. Hillol Kargupta
University of Maryland Baltimore County and President, AGNIK, LLC., USA

Agent Mediated Knowledge Management (AMKM) (Tutorial notes)
Dr. Virginia Dignum
Utrecht Universit, The Netherlands

Distributed Constraint Reasoning: A Paradigm for Effective Coordination in Multiagent Systems (Tutorial notes)
Makoto Yokoo, Kyushu University, Japan
Joerg Denzinger, University of Calgary, Canada
Marius Silaghi, Florida Tech, USA
Adrian Petcu, Swiss Federal Inst. of Tech.

Tutorial Chair

Pawan Lingras Saint Mary's University, Canada
(E-mail: pawan@cs.smu.ca or Pawan.Lingras@smu.ca)

Tutorial Abstracts

Web-based Support Systems

Dr. JingTao Yao
University of Regina, Canada

Abstract

Moving support systems online is an increasing trend in Web Intelligence research. Web-based Support Systems (WSS) are an emerging multidisciplinary research area that studies the support of human activities with the Web as the common platform, medium and interface. One of the goals of building WSS is to extend the human physical limitation of information processing in the information age.

Research on WSS is motivated by the challenges and opportunities arising from the Internet and the Web. The availability, accessibility and flexibility of information and the tools to access this information lead to many opportunities. However, there are also many challenges. We have to deal with more complex tasks, as there are more demands for quality and productivity. WSS are a natural evolution of the studies on various computerized support systems such as Decision Support Systems, Computer Aided Design, and Computer Aided Software Engineering. The recent advancements of computer and Web technologies make the implementation of WSS feasible. It is rare to see a system without some type of Web interaction.

In this tutorial, I will discuss fundamental issues of WSS, a framework for WSS, and current research on WSS. A key issue of WSS research is to identify both domain independent and dependent activities before selecting suitable computer and Web technologies to support them. I will also briefly introduce specific types of WSS: Web-based research support systems and Web-based information retrieval support systems. As an example, I will use Web of Science, a well-known citation database, to demonstrate the effectiveness of this research domain.

Bio

Dr. Yao is an Associate Professor of Computer Science at the University of Regina. Before arriving Canada, he taught at Massey University, New Zealand, the National University of Singapore, Singapore, and Xi'an Jiaotong University, China. He received his Ph.D. degree at the National University of Singapore and did a B.Eng. degree and an M.Sc. degree at Xi'an Jiaotong University. His research interests include soft computing, data mining, forecasting, neural networks, computational finance, electronic commerce, Web intelligence, and Web-based support systems. He has published more than 60 papers in these areas. Dr. Yao chaired three international workshops on Web-based Support Systems.

top

Web and Text Mining for Opinion/Trend Analysis

Dr. Lipika Dey
Tata Consultancy Services, India

Abstract

The web today has emerged into a gothic heterogeneous structure that acts like a community black-board which contains information about requirements and availability of knowledge on all sundry topics, cutting across geographical boundaries. It comprises structured, semi-structured, and unstructured documents which in turn could be controlled or uncontrolled in terms of the contents contained in them. The snapshot of the Net at an instant of time effectively represents the collective brain map of a large community across geographical and social barriers. As a consequence it can play a major role in determining as well as influencing collective interests, trends and opinions of the masses through news sites, social networks, blogs etc. provided appropriate knowledge discovery techniques are applied to do so. While opinion mining from text documents aims at understating users' collective opinions about an entity as expressed in the web, and can be broadly categorized as negative or positive, trend analysis aims at studying the variation of contents expressed over time.

Knowledge discovery from unstructured text employs a large array of analytical techniques ranging from social science of human behavioral patterns to statistics, web-structure analysis, Natural Language Processing, Data Mining and Knowledge Discovery etc. are adopted to discover do this. This tutorial is aimed at introducing the web and text mining techniques that are adopted for opinion mining and trend-analysis from unstructured content of the web. It will begin with a discussion on link-based text processing techniques that can generate concept maps which provide effective abstractions of the web structure and its content for further analysis. Link-based analysis can also be used for analysis of collective-user behavioral patterns. Thereafter, the discussion will be centered on tools and techniques used mining information from text documents. Text mining involves two key tasks - text information extraction and analysis. Several Natural Language Processing (NLP) tools and techniques that are employed for text information extraction will be discussed. Besides taggers, parsers and dependency analysis which play key roles in understanding textual content, mechanisms for identifying events, opinions, emotions, temporal information etc. will be presented in this tutorial. Several statistical, data mining and machine learning algorithms that are applied to do these in addition to NLP and ontology-based reasoning will be discussed. Opinion extraction through text mining involves collating data from multiple sources and finally come up with conclusions that may not be explicit in the extracted data but can be inferred from the data using inference engines. Such information has a huge commercial impact and can play a crucial role in influencing business policies and decisions. Opinion and trend analysis are sought after by product companies to get feedback about their products, by company CEOs to gauge their market-rating, by web-site designers to attract clients and of-course by all prospective customers to get a feedback on any product they decide to buy.

Bio

Dr. Lipika Dey is engaged as a Consultant with Innovation Labs, Tata Consultancy Services, a premier consulting organization of India. Her research is focused towards knowledge discovery from structured and unstructured documents, text mining, and machine learning. Prior to her current assignment, Lipika was in the faculty of Dept. of Mathematics, I.I.T. Delhi from 1995 to 2006. She has also worked as a Research Scientist in Webaroo Technology India Pvt Ltd., where she was involved in conceptualizing and designing Text Extraction based products for the mobile web. Her primary interest lies in developing efficient mechanisms for extracting information components from unstructured text documents, and exploit these for predictive analysis. Designing novel visualization techniques for easy dissemination and assimilation of textual information are also of key interest to her.

She, along with her colleagues has worked extensively in Bio-medical document processing for intelligent information retrieval. They have worked towards utilizing entity-relationship based information components to summarize document collections for easy comprehension of web users. Currently, she is engaged in developing techniques for identifying business process strategies from enterprise document collections. Machine learning techniques are adopted to work seamlessly with information collected from structured and unstructured information sources for various domains like insurance sectors, legal documents and so on. As a faculty member in IIT in the past, she has guided more than 50 Master’s theses and two PhD theses. She has also worked in designing supervised and unsupervised learning mechanisms for analyzing categorical data stored in structured domain repositories. She specializes in application of soft-computing techniques towards various knowledge-discovery problems, both from structured and unstructured texts. Her extra-academic interests include creative writing and dramatics, which she actively pursues whenever time permits.

top

Agent-Mining Interaction and Integration

Dr. Longbing Cao
University of Technology, Sydney, Australia

Abstract

In the last decade, agents and data mining have emerged as two of most vivacious areas in information technology field. The complementary nature of both areas foreshows an emergent trend that is the increasing interaction between them. As a result, both sides obtain great enhancement in terms of solving the existing individual challenges and innovating new research and development opportunities. The interaction will foster prospects towards the advancements of next-generation intelligent technologies, processing, systems and services from theoretical, technological and practical perspectives. This tutorial will address the state of art of aims, challenges, technologies and practices in agent-mining interaction and integration. In particular, we will address the following key issues:

  • Evolution of the agent and data mining interaction and integration;
  • Common issues in agent and data mining;
  • Key challenges and benefits in agent-mining interaction and integration,
  • Technologies and practices of agent driven data mining;
  • Technologies and practices of data mining driven agents and multi-agent systems;
  • Several real world applications and systems.
It will show that some of challenges in either community may be effectively or efficiently tackled through agent-mining interaction and integration. The tutorial targets academics, industrial researchers, and developers interested in research and development in autonomous agents and multiagent systems, data mining and knowledge discovery, as well as developing the next generation intelligent technologies, systems, and services.

Bio

Dr Longbing CAO is a Senior Lecturer of the Faculty of Information Technology, University of Technology Sydney, Australia. He is one of pioneer researchers in initiating and promoting the emergent area of agent-mining interaction and integration. He was program co-chairs of three international workshops on this topic. He organized the first special issue and a website on this issue. He has developed online systems not only for research in agents, data mining and agent-mining interaction, but also practical applications such as financial and economic trading services.

top

Trust Mechanisms for Agent Systems

Dr. Sandip Sen
University of Tulsa, USA

Abstract

As agents interact in open environments, the need for mechanisms that allow the representation of and reasoning with the trust and reputation of other agents become critical. Application domains including e-commerce, e-services, supply chains, social networks, peer-to-peer systems, etc. require participants to repeatedly interact with and rely on other entities in the environment. Whereas economic mechanisms like auctions and negotiation protocols that require exchange of monetary units are valuable tools for one-off interactions, sustained or repeated agent interactions can be better supported by social mechanisms based on trust. The multiagent systems research community have studied various trust, reputation, and referral frameworks and models that provide the necessary building blocks to develop applications with embedded trust models and reasoning procedures. In this short tutorial we will overview some of the better known model with emphasis on diversity of techniques to cover maximal potential application scenarios.

Bio

Sandip Sen is a Professor of Computer Science in the University of Tulsa with primary research interests in multiagent systems, machine learning, and genetic algorithms. He completed his PhD in the area of intelligent, distributed scheduling from the University of Michigan in December, 1993. He has authored approximately 200 papers in workshops, conferences, and journals in several areas of artificial intelligence. In 1997 he received the prestigious CAREER award given to outstanding young faculty by the National Science Foundation. He has served on the program committees of most major national and international conferences in the field of intelligent agents including AAAI, IJCAI, ICMAS, AA, AAMAS, ICGA, etc. He was the co-chair of the Program Committee of the 5th International Conference on Autonomous Agents held in Montreal Canada in 2001. He regularly reviews papers for major AI journals and serves on the panels of the National Science Foundation for evaluating agent systems related projects. He has chaired multiple workshops and symposia on agent learning and reasoning. He has presented several tutorials on multiagent systems in association with the leading international conferences on autonomous agents and multiagent systems.

top

Peer-to-Peer Distributed Data Mining for Multi-Agent Applications

Dr. Hillol Kargupta
University of Maryland Baltimore County and President, AGNIK, LLC., USA

Abstract

This tutorial will present an overview of the emerging Peer-to-Peer (P2P) data mining technology for multi-agent applications. It will first discuss different issues relevant to P2P computing and identify the challenges for building data mining applications in such environments. This tutorial will then review algorithmic foundations for P2P distributed data mining and monitoring. It will discuss several recently developed approximate and exact P2P algorithms for distributed clustering, inferencing, predictive modeling and outlier detection, among others. The tutorial will also discuss several applications, including one on multi-agent P2P client-side web mining. In addition, the tutorial will point the attendees toward publicly available resources and literature.

Bio

Hillol Kargupta is an Associate Professor in the Department of Computer Science and Electrical Engineering, University of Maryland, Baltimore County. He received the PhD degree in computer science from the University of Illinois at Urbana-Champaign in 1996. He is also a co-founder of Agnik LLC, a data analytics company for distributed, mobile, and embedded environments. His research interests include mobile and distributed data mining and computation in biological process of gene expression. Dr. Kargupta won a US National Science Foundation CAREER award in 2001 for his research on ubiquitous and distributed data mining. He along with his coauthors received the best paper award at the 2003 IEEE International Conference on Data Mining for a paper on privacy-preserving data mining. He won the 2000 TRW Foundation Award and the 1997 Los Alamos Award for Outstanding Technical Achievement. His research has been funded by the US National Science Foundation, US Air Force, Department of Homeland Security, NASA, and various other organizations. He has published more than ninety peer-reviewed articles in journals, conferences, and books. He has co-edited two books: Advances in Distributed and Parallel Knowledge Discovery, AAAI/MIT Press, and Data Mining: Next Generation Challenges and Future Directions, AAAI/MIT Press. He is an associate editor of the IEEE Transactions on Knowledge and Data Engineering, IEEE Transactions on Systems, Man, and Cybernetics, Part B and Statistical Analysis and Data Mining Journal. He was the program-co-chair of the 2005 SIAM Data Mining Conference, Program vice-chair of 2005 PKDD Conference, Program vice-chair of 2005 IEEE International Data Mining Conference, Program Vice Chair for 2005 Euro-PAR Conference, Associate General Chair of the 2003 ACM SIGKDD Conference, and chair of the 2002 NSF Next Generation Data Mining Workshop among others. He regularly serves in the organizing and program committee of many data mining conferences. More information about him can be found at http://www.cs.umbc.edu/~hillol.

top

Tutorial on Agent Mediated Knowledge Management

Dr. Virginia Dignum
Utrecht Universit, The Netherlands

Abstract

Agent-mediated Knowledge Manage¬ment (AMKM) aims at developing comprehensive solutions to knowledge management problems in real-world environments. Agent-based approaches can deal with collective aspects of the domain in an attempt to cope with the conflict between desired order and actual behavior in dynamic environments. Inherent to AMKM is a social layer, which structures the society of agents by defining specific roles and possible interactions between them. An obvious consequence of this social perspective is that communication must be considered as a first class concept in AMKM. In heterogeneous KM systems, knowledge sharing is hampered by the lack of common ontologies. Therefore, adequate support for ontology matching and meaning negotiation is of great importance to AMKM.

AMKM is an interdisciplinary research area as it combines predominant trends in business, such as knowledge management, with state of the art AI techniques, such as MAS, text classification and ontologies. Typical characteristics of KM environments are: manifold logically and physically dispersed actors and knowledge sources, different degrees of formalization of knowledge, different kinds of (web-based) services and (legacy) systems, and possible conflicts between individual and global (group or organizational) goals.

This 2-hour tutorial will discuss methodological, technical and application aspects of AMKM. In particular, we will discuss the following topics:

  • Supporting KM with Agent Technology
  • KM issues in multi-agent systems
  • Specification and design of AMKM Systems
  • Methodologies for AMKM
  • Knowledge representation in AMKM
  • Applications of AMKM

Bio

Virginia Dignum is assistant professor in the Cognition and Communication group of the Institute for Information and Computing Science, Utrecht University. She has a PhD in Computer Science from Utrecht University, the Netherlands. She worked in industry for more than 12 years and has vast experience in consultancy and system development in the areas of expert systems and knowledge management. Her research focuses on agent based models of organizations, in particular in the dynamic aspects of organizations, and the applicability of agent organizations to support knowledge creation, sha¬ring and representation in distributed, heterogeneous environments. She was the initiator of the workshop series on Agent-Mediated Knowledge Management, has edited a book on AMKM and written several articles on this subject. She organized several international workshops, including the related workshops listed below, was local co-organizer and treasurer of the 4th International Conference in Autonomous Agents and Multi-Agent Systems (AAMAS 2005), and has served as PC member in many international conferences, journals and workshops. She has participated in several national and EU-projects, and has more than 50 published articles in books, conference proceedings and journals.

top

Distributed Constraint Reasoning - A Paradigm for Effective Coordination in Multiagent Systems

Makoto Yokoo , Kyushu University, Japan
Joerg Denzinger , University of Calgary, Canada
Marius Silaghi , Florida Tech, USA
Adrian Petcu , Swiss Federal Inst. of Tech.

Abstract

Constraint satisfaction/optimization is a powerful paradigm for solving numerous practical problems like planning, scheduling and resource allocation. There exists a large body of work tackling such problems in a centralized setting, with centralized algorithms. However, many real problems are naturally distributed between a set of agents, each one holding its own subproblem, which has interdependencies with (some of) its peers' subproblems. The Distributed Constraint Optimization Problem (DCOP) is a framework that has thus recently emerged as a promising approach to addressing complex coordination problems in Multiagent Systems.

In DCOP, the agents communicate through message exchange to find the optimal solution to the overall optimization problem. Key issues are efficiency (minimizing communication and memory requirements), coping with system dynamics (problems can change at runtime), privacy (leaking as little private constraints and valuations as possible) and incentives (designing algorithms that ensure honest behavior from self-interested agents).

This tutorial will provide an unified view on Distributed Constraint Reasoning, introducing distributed constraint reasoning systems as semi-cooperative multi-agent systems and concentrating on the issues of efficiency, privacy, and dynamic systems. The general ideas behind the known distributed constraint reasoning systems are presented within this multi-agent framework. For each approach, the requirements, limitations, advantages and disadvantages of the different categories will be discussed.

The tutorial is structured as follows: In the first part of this tutorial we introduce the DCOP framework, discuss motivation, and present an example application mapped as a DCOP.

In the second and the third parts we present the two main classes of algorithms for DCOP: part 2 is dedicated to search algorithms, and part 3 to dynamic programming algorithms. From each class, the most representative algorithms are described. Relationships are explained, and strengths and weaknesses are compared.

In the last part of this tutorial we discuss privacy-preserving DCOP algorithms for semi-cooperative multiagent systems.

Prerequisite knowledge:
The tutorial is targeted at the general AI audience, both academic and industrial. It requires only a basic knowledge in standard algorithmic schemes, like branch-and-bound or dynamic programming.

Bio

Joerg Denzinger
Joerg Denzinger is an Associate Professor of Computer Science at the University of Calgary for Artificial Intelligence and Multi-Agent Systems. He has a PhD from the University of Kaiserslautern (1993), where he also did his Habilitation (2000). His research interests include distributed knowledge-based search, opponent modeling, distributed data mining, learning in multi-agent systems, and testing of multi-agent systems.

Previous tutorials:

  • AAMAS-2004 tutorial on Distributed Constraint Reasoning, New York
  • IJCAI-2003 tutorial on Distributed Constraint Reasoning, Acapulco
  • IJCAI-2001 tutorial on Distributed Knowledge-based Search, Seattle

Marius Silaghi
Marius Silaghi got his PhD in 2002 from the Swiss Federal Institute of Technology at Lausanne (EPFL) with a thesis about privacy for Distributed Constraint Satisfaction. Since 2002 he is assistant professor in the Computer Sciences department at the Florida Institute of Technology. His research and teaching is centered on Distributed Constraint Reasoning, and has helped organize several workshops and special tracks on this topic.

Previous tutorials :

  • Tutorial on Distributed Constraint Reasoning, IJCAI 2003, Acapulco Mexico
  • Tutorial on Distributed Constraint Reasoning, AAMAS 2004, Columbia Univ NY
  • Tutorial on Distributed Constraint Reasoning, IJCAI 2005, Edinburgh

Adrian Petcu
Adrian Petcu is a PhD student in the Artificial Intelligence Laboratory from the Swiss Federal Institute of Technology at Lausanne (EPFL) since 2002. He will graduate in September 2007, with a thesis about Distributed Constraint Optimization. The focus of his research is on efficient algorithms for distributed constraint optimization, but his broader interests also touch upon aspects of dynamic problem solving, game theory and its applications to settings with self-interested agents, and privacy. He has published extensively on these topics in well known conferences like IJCAI, AAAI and CP.

He was a co-organizer of several DCR-related workshops, and the chair of DCR'07, held in conjunction with IJCAI'07.

Last update: 28 Jul 2011 top

Web Intelligence Consortium