K2LAB - Knowledge Engineering and Communication

This document is a preliminary version and does not contain the full or final text. Mon Oct 10 11:40:26 MET 1994
calle@dsv.su.se

Scope

The overall objective for the laboratory is to tune the design and use of computer-based systems to the human user's cognitive capabilities and to let models and theories of human cognitive processes inspire design principles for computer-based systems.

Cognitive capabilities include our conception of the world, how we acquire and apply knowledge and skills (learning, memory and problem solving) and how we communicate and act in interplay with our environment.

The goal is to develop formal models for representation, problem solving, adaptation (learning) and communication which can serve as the basis for realization of computerized support systems within application areas like decision making, design, validation, monitoring and diagnosis, communication (electronic mail and tele- and video conferences), education (both at universities, schools and training-services) and the dissemination of current and rapidly changing information (legislation, environmental, medical and technical expertise).

We need to combine techniques from areas like programming methodology (PM), artificial intelligence (AI), human-machine interaction (HCI), computer supported cooperative work (CSCW) and natural language processing technology (NLP). The research methodology should include task analysis and user oriented design, formal modelling and algorithm design, exploratory implementations and analytic and empirical validation.


Table of Contents


Research Topics

Knowledge Representation

Text missing.

Knowledge Engineering

This area is concerned with research topics related to the whole development cycle of knowledge-based systems systems, in particular with the question of how to support the activities of eliciting and representing knowledge in such systems. The approach is to develop reusable problem-solving methods and domain models (reusable knowledge modelling components) for industrially important knowledge-based systems applications. These includes technical diagnose, configuration, prediction planning and design support. The major goal within the field is to create a level of abstraction where automated problemsolving can be understood, and thus controlled, by domain experts and users, allowing them to take active part in the development process.

Machine Learning

The main question with which this area is concerned is how to create adaptive systems by integration of learning algorithms as components. This involves technical work on representation, reasoning and learning algorithms as well a studies of the cognitive relevance of the proposed solutions and the application of solutions in realistic applications.

Of secondary importance is the use of learning algorithms as an aid for knowledge acquisition, inducing elements of a theory from observed cases. The research focusses on symbolic machine learning techniques. The emphasis is on the revision of theories in Horn clause logic by combination of techniques from the specialized areas of Inductive Logic Programming and Explanation-based learning. Another important topic is conceptual clustering applied in real-time scenarios. To a smaller extent nonsymbolic approaches have been used (genetic algorithms).

An important methododological foundation is the application of formal logic for representation as well as problem solving and logic programming (LP) as a basis for computation. Within this general scheme there is a spectrum of approaches from Horn clause logic, via extensions of logic programming to standard FOL and to non-standard logics.

Distributed AI and Federated Deductive Databases

The main research question is to extend knowledge-based methods to deal with large amounts of real-world information. Rather than modelling information statically as a consistent set of atomic relational positive facts, the store of knowledge or information is supposed to be distributed over a network of nodes in less than perfect communication with each other. At each node, the store of information can be fluctuating in time, incomplete or inconsistent and the individual pieces of information can be logically complex, negative or indefinite.

The proposed methods are techniques from mathematical logic, the theory of logic programming and the study of formal representation of non-monotonic reasoning. To model interaction between nodes, axiomatic modal logic as well as temporal logic can be applied. To represent the state at each node a particular approach to non-monotonic reasoning based on a a prolog-type language with both classical negation and negation as failure is used (Busch, Boman).

Reasoning under Uncertainty and Decision Theory

This research concerns the area of mechanizing decision analysis. The work includes both mathematical models, psychological studies and models, the design of computer support systems and practical applications. One approach is the model for decision analysis proposed by Malmnäs. A particular system based upon this model includes well-founded routines that enable the decision maker to work with a vague and numerically imprecise basis for decision and, despite this, reach a conclusive result. Particular problems to be investigated in this context are to develop:
  1. suitable algorithms for evaluating decision problems,
  2. methodologies and algorithms for taking account of different kinds of probability and utility distributions according to a set of criteria,
  3. the interpretation of evaluation of fixed alternatives over vague domains, and
  4. methodologies and algorithms for generating optimal alternatives out of an unspecified set of alternatives.
It is also important to study to what extent an automated system would be able to model different behavior in decision situations (Malmnäs, Ekenberg, Montgomery).

Human-Computer-Interaction

Text missing.

Computer Mediated Communication and Computer Supported Cooperative Work

The areas of computer-mediated communication (CMC) and computer supported cooperative work (CSCW) are closely related and share the objective to develop computer-based aids for cooperation and communication in groups of individuals. The term CMC is mainly used for applications where the participants do not need to be connected all at the same time, while CSCW also includes same-time applications.

Among applications should be mentioned electronic mail, computer conferencing and bulletin board systems, computer support for particular cooperative processes like idea collection, text editing, voting and question answering. Typical research issues for these kinds of systems are: to develop good interfaces, studies of effects on people and organizations, software support systems, hypermedia databases, distributed protocols, economical and market issues and standardization processes.

Current Research Topics

Communication Models

Mikael Kindborg, Robert Ramberg, (Klas Karlgren) and Carl Gustaf Jansson.

Intelligent Interfaces

Klas Karlgren, Ann Lantz, Fredrik Kilander, Carl Gustaf Jansson (and Jacob Palme).

The focus of research lies in marrying user-centered design methods and user-controlled interface designs, with the advanced communication strategies from intelligent interfaces. This is achieved by focussing on the development of adaptive interfaces. The same emphasis is put on finding mechanisms (models and algorithms) for user and context adaptation as well as methods for requirement engineering in this area. Mechanisms of particular interest are user modelling techniques, planning and plan inference algorithms.

Multi-Agent Architectures in a Logic Framework

Magnus Boman, Douglas Busch (and Love Ekenberg).

Integrated Theory Revision Systems in a Logic Programming Framework

Peter Idestam-Almquist, Henrik Boström, Carl Gustaf Jansson, and Lars Asker.

Decision Modelling

Love Ekenberg, Mats Danielsson (and Magnus Boman).

User-Adapted Tele-Conferencing Systems

Jacob Palme, Torgny Tholérus and Tarja Lintunen.

Computer-Based Learning Technologies

Anita Kollerbaur, Carl Gustaf Jansson and Douglas Busch.

Staff

Research management

Carl Gustaf Jansson, acting professor, PhD, (coordinator), artificial intelligence, knowledge representation, machine learning, intelligent interfaces and computer supported education.

Jacob Palme, associate professor, Techn. Lic. computer supported cooperative work and computer mediated communication.

Senior researchers, lecturers and postdocs

Lars Asker, lecturer, PhD., artificial intelligence, machine learning, speedup-learning.

Henrik Boström, lecturer, PhD., artificial intelligence, machine learning, logic programming, inductive logic programming.

Magnus Boman, lecturer, PhD., knowledge representation, modelling of autonomous deductive systems, multi-agent architectures and decision analysis.

Douglas Busch, associate professor, PhD., knowledge representation in logic, multi-agent architectures and logic programming.

Love Ekenberg, lecturer, PhD., decision modelling, reasoning under uncertainty.

Peter Idestam Almquist, PhD., artificial intelligence, machine learning, logic programming, inductive logic programming.

Fredrik Kilander, PhD., artificial intelligence, machine learning.

Harald Kjellin, PhD., knowledge-based systems, knowledge representation and knowledge acquisition.

Anita Kollerbaur, senior lecturer, Fil. Lic., human-computer-interaction and computer-based education.

Graduate students with graduate positions/research engineers

Bassam Michel El Khouri, Fil. lic., artificial intelligence, machine learning and statistics, reasoning under uncertainty.

Peter Holm, Fil. lic., knowledge-based systems, cognitive science, social and organizational aspects of computerization.

Klas Karlgren BA., human-computer-interaction, cognitive psychology.

Hercules Dalianis, Techn. lic., natural language processing, text generation.

Ann Lantz, Fil. Lic., human-computer-interaction, cognitive psychology.

Robert Ramberg, Fil. Lic., human-computer-interaction, learning, cognitive psychology.

Mats Danielsson, MSc., decision modelling, reasoning under uncertainty.

Tarja Lintunen BA., computer mediated communication.

Åsa Rudström, BA., artificial intelligence, knowledge representation, machine learning.

Torgny Tholerus, BA., computer mediated communication.

Pierre Wijkman, MSc., artificial intelligence, machine learning, genetic algorithms.

Associated graduate students

At SICS:
in natural language processing
in knowledge-based systems
in human-computer interaction
At the Sweden Mid University: Åke Malmberg.

At Naturens Hus: Mikael Kindborg, Fil. Lic., human-machine interaction, multimedia technology.

Associated advisors

Jan Olsson, PhD., SICS.

Per Erik Malmnäs, Doc, Philosophy, Stockholm University.

Ywonne Waern, Prof., TEMA Communication, Linköping University.


Results

The Lab has produced 10 PhDs and 18 Licentiates since 1990, a majority of the PhDs during 1993-1994.

Algorithms and prototype systems in the area of Machine Learning:


Current Projects and Research Activities

ACTA

Objectives
To give technical contribution to symbolic machine learning techniques.
Administrative facts
Funding: NUTEK under the grant XXXXXXX.
Duration: six years 1987-1993.
Staff: Carl Gustaf Jansson, Peter Idestam Almquist, Henrik Boström, Lars Asker, Fredrik Kilander, Bassam Michel El Khouri, Pierre Wijkman and Åsa Rudström.
Collaboration: European partners within the COST 13 Acqusition and Learning program.
Main Results
Five PhD theses completed during 1993/94. An algorithm for generalization of hornclauses under implication (Summer 1993). An EBG algorithm that guarantees nonredundancy (Summer 1993). An enhanced conceptual clustering algorithm that is robust with respect to concept drift (Autumn 1993).
Plans
The project is finished in 1993 but as a side effect another two PhD theses and one licentiate thesis will be finished in the autumn of 1994

Inductive Machine Learning Techniques

Objectives
To give technical contribution to inductive symbolic machine learning techniques in a logic programming project.
Administrative facts
Funding: TFR under the grant XXXXXXX.
Duration: three years 1993-1996.
Staff: Carl Gustaf Jansson, Peter Idestam Almquist.
Collaboration: European partners within the ILP project.
Main Results
An enhanced algorithm for generalization under implication.

Applications of Machine Learning Techniques

Administrative facts
Funding: NUTEK under the grant XXXXXXX.
Duration: Three years 1993-1996.
Staff: Carl Gustaf Jansson, Lars Asker and Robert Engels.
Collaboration: The Högdalen energy plant, Ångpanneföreningen.
Main Results
A prototype system for optimizing a garbage burning energy plant using two alternative inductive learning algorithms (Summer 1994).

PUSH - Plan and User Sensitive Help

Administrative facts
Funding: NUTEK under the grant XXXXXXX.
Duration: Three years 1993-1996.
Staff: Carl Gustaf Jansson, Klas Karlgren, Nils Dahlbäck, (Catriona MacDermid).
Collaboration: SICS and Ellemtel.
Main Results
A help system for a software engineering method and its toolbox (Summer 1994).

AUGMENTation of expertise

Administrative facts
Funding: NUTEK under the grant XXXXXXX.
Duration: One year 1993-1994, possible cont. for 94.
Staff: Carl Gustaf Jansson, Klas Karlgren, Robert Ramberg.
Collaboration: Telia

Communicating Energy and Environmental Expertise

Administrative facts
Funding: NUTEK under the grant XXXXXXX.
Duration: One year 1993-1994.
Staff: Carl Gustaf Jansson, Joanna Dickinsson and Harald Kjellin.
Collaboration: Ångpanneföreningen.
Main Results
A prestudy of how a distributed multimedia system can assist in planning for energy consultants (Summer 1994).
Plans
An extension of the project is planned involving a field study.

Federated Information Systems Technology

Administrative facts
Funding: graduate position and other faculty funding.
Duration: Three years 1990-1993.
Staff: Magnus Boman and Douglas Busch.
Collaboration: The COMPULOG NETWORK.

Decision Modelling for Vague Domains

Administrative facts
Funding: graduate position and other faculty funding.
Duration: Three years 1991-1994.
Staff: Love Ekenberg and Mats Danielsson.
Main Results
A PhD thesis in april 1994.

INTFILTER - Intelligent Filtering of Usenet News and Mail

Administrative facts
Funding: NUTEK and AMFO under the grant XXXXXXX.
Duration: Three years 1993-1996.
Staff: Jacob Palme, Fredrik Kilander, Ann Lantz, Daniel Pargman, Carl Gustaf Jansson and Ywonne Waern.

User-Adapted Tele-Conferencing Systems

Administrative facts
Funding: NUTEK and AMFO under the grant XXXXXXX.
Duration: three years 1993-1996.
Staff: Jacob Palme, Tarja Lintunen.

Multimedia Interfaces to Tele-Conferencing Systems

Administrative facts
Funding: NUTEK and AMFO under the grant XXXXXXX.
Duration: One years 1993-1996.
Staff: Jacob Palme, Anita Kollerbaur and Torgny Tholerus.

Collaboration

Text missing.