This paper presents a Knowledge Representation Object Language (KROL) on top of Prolog. KROL aimed at providing the ability to develop second-generation expert systems. The main aspects of KROL include multi-paradigm knowledge representation (first-order predicate logic, objects, rules), inference mechanisms at different levels of granularity, explanation facility, object-oriented database management module, and user-friendly interface. KROL has sufficient expressive power to be used in applying demanding knowledge-based modeling methodologies, such as KADS and Generic Task, which are the major landmarks of the second-generation expert systems technology. Four successful agricultural expert systems were developed in the last six years using KROL. To demonstrate the language capabilities, we present an example of disorder diagnosis. 2 1.

Report on the workshop on Knowledge Level Models of Machine Learning that was organized in the context of the second series of MLNet familiarization workshop, 9-10 April 1994, Catania, Italy. Organization: Walter Van de Velde Artificial Intelligence Laboratory Vrije Universiteit Brussel Pleinlaan 2, B-1050 Brussels, Belgium Email: walter@arti.vub.ac.be 1 Topic Description The aim of this workshop was to discuss knowledge level modeling applied to machine learning systems and algorithms. An important distinction in current expert systems research is the one between knowledge level and symbol level [Newell, 1982]. Systems can be described at either of these levels. Briefly stated, a knowledge level description emphasizes the knowledge contents of a system (e.g. goals, actions and knowledge used in a rational way) whereas the symbol level describes its computational realization (in terms of representations and inference mechanisms). There is a consensus that modeling at the knowledge lev...

This document provides a formal model for Knowledge-Based Systems(KBS). It outlines the formal modelling of the different components of a Rule-Based System (RBS). Initially abstract, the model becomes more concrete in a sequence of refinements., naturally evolving to address such aspects of a RBS like knowledge representation, mode of reasoning, consistency. An example of instantiation of the general model in order to provide a particular RBS is also shown. The problem dealt with is about the choice of satisfying insurance policies for a given customer. An informal description of this application domain is given in an appendix. Souleymane Koussoub'e is a fellow of UNU/IIST from February to May 1997. He got his Doctorate in Computer Science from the University Paul Sabatier of Toulouse (France). He is on leave from the African's Institute of Computer Science (Libreville), where he is lecturer. His research interests include Decision Support Systems, Formal Development of Software Syst...

Knowledge Engineering (KE) and Software Engineering (SE) have similar goals: developing methods, techniques, and tools for the building process of either knowledge based systems (kbs) or (complex) traditional software. Due to this kinship it seems obvious to analyse to which extent both areas, KE and SE, can benefit from each other. In this paper we want to argue from the KE point of view. The main paradigm in KE has switched from a transfer point of view to a modeling point of view. So a lot of emphasis has been put on the development of notions and methods for building and structuring models, which capture different results of the development process. Much effort has been investigated into the analysis and development of reusable components: problem solving methods describing the dynamic behaviour of kbs and ontologies defining the vocabulary and structure of (domain) models. We describe specific approaches in KE in some detail which exploit these ideas in different ways: Role-Limiting Methods, KADS and MIKE. Finally Life Cycle Models, Non Functional Requirements and Transformational Development are discussed as areas where we think KE might profit from research and experiences made in SE.

avail. Merci pour son soutien et ses remarques toujours pertinentes et constructives. Merci Franois Lecordix, de l'IGN, pour son travail sur le formidable outil d'exprimentation PlaGe, pour son travail ralis en gnralisation du rseau routier, pour sa chasse au bugs, pour ses ides qui ont fortement influences GALBE, et pour avoir permis que cet algorithme soit utilis dans les services de production de l'IGN. Merci Jean-Gabriel Ganascia pour m'avoir accueilli au sein de l'quipe ACASA du LIP6. Merci Grald Weger, mon premier professeur de cartographie l'ENSG, pour m'avoir transmis sa passion de ce domaine, et pour avoir accept de commenter certains des rsultats de cette thse durant sa retraite. Merci Nicolas Regnauld, de l'universit d'Edimbourgh, pour m'avoir fourni de nombreux exemples ncessaires mon travail et pour son aide de manire gnrale. Merci Lorenza Saitta, spcialiste de l'apprentissage automatique, pour avoir accept de poser son regard trs pertinent sur mes travaux. Merci R

r permis que cet algorithme soit utilis dans les services de production de l'IGN. Merci Jean-Gabriel Ganascia pour m'avoir accueilli au sein de l'quipe ACASA du LIP6. Merci Grald Weger, mon premier professeur de cartographie l'ENSG, pour m'avoir transmis sa passion de ce domaine, et pour avoir accept de commenter certains des rsultats de cette thse durant sa retraite. Merci Nicolas Regnauld, de l'universit d'Edimbourgh, pour m'avoir fourni de nombreux exemples ncessaires mon travail et pour son aide de manire gnrale. Merci Lorenza Saitta, spcialiste de l'apprentissage automatique, pour avoir accept de poser son regard trs pertinent sur mes travaux. Merci Robert Weibel, directeur du groupe de travail de l'ACI en gnralisation, pour avoir accept d'valuer cette thse inspire de ses premiers travaux en apprentissage automatique pour la gnralisation cartographique. Merci Bndicte Bucher la KADSienne, pour ses conseils dans ce domaine, et son soutien permanent. Merci aux membres de l'qui

Abstract — The classification methods applied in the objectoriented image analysis approach are often based on the use of domain knowledge. A key issue in this approach is the acquisition of this knowledge which is generally implicit and not formalized. In this paper, we examine the possibilities of using genetic programming for the automatic extraction of classification rules from urban remotely sensed data. The method proposed is composed of several steps: segmentation, feature extraction, selection of training sets, acquisition of rules, classification. Features related to the spectral, spatial and contextual properties of the objects are used in the classification procedure. Experiments are made on a Quickbird MS image. The quality of the results shows the effectiveness of the proposed genetic classifier in the object-oriented, knowledge-based approach. Keywords- genetic classifier, knowledge-based classification method, object-oriented approach. I.