A number of groups developing knowledge-based systems have found (or at least posited) that the design and representation of a limitative set of concepts and relations, a so-called ontology, can contribute to sharing and reusing knowledge bases. However, very few descriptions of implemented ontologies have appeared in the literature. No comparison of competing proposals is available, let alone an empirical determination of the benefits of using an ontology. There is no accepted method for designing and building such ontologies, nor is it clear how ontologies can best be evaluated.

Assuring the reliability of knowledge-based systems has become an important issue in the development of the knowledge engineering discipline. There has been a workshop devoted to these topics at most of the major AI conferences (IJCAI, AAAI and ECAI) for the last five years, and the 1994 European Conference on Artificial Intelligence (ECAI-94) in Amsterdam was no exception. The focus of the meeting was on validation techniques for KBS, where validation is defined as the process of determining if a KBS meets its users' requirements; implicitly, validation includes verification, which is the process of determining if a KBS has been constructed to comply with certain formally-specified properties, such as consistency and irredundancy. The Amsterdam workshop was an intimate meeting, and the fifteen attendees were predominantly from European institutions. In spite of—or perhaps because of—this intimacy, the workshop succeeded in highlighting many of the significant trends and issues within its area of concern. The purpose of this short article is to review the trends and issues in question, drawing upon the contributions made during the workshop.

Structured objects are items with defined properties that are to be represented in a computer system. Research in Knowledge Representation (KR) and in Database Design (DB) has produced languages for describing structured objects. Although different in the particular means for defining properties, both areas share the goal of representing a part of the world in a structured way. Moreover, the rise of object-centred formalisms in the last decade has significantly influenced the convergence of languages.

“Integration of Machine Learning and Knowledge Acquisition” may be a surprising title for an ECAI-94 workshop, since most machine learning (ML) systems are intended for knowledge acquisition (KA). So what seems problematic about integrating ML and KA? The answer lies in the difference between the approaches developed by what is referred to as ML and KA research. Apart from sonic major exceptions, such as learning apprentice tools (Mitchell et al., 1989), or libraries like the Machine Learning Toolbox (MLT Consortium, 1993), most ML algorithms have been described without any characterization in terms of real application needs, in terms of what they could be effectively useful for. Although ML methods have been applied to “real world” problems few general and reusable conclusions have been drawn from these knowledge acquisition experiments. As ML techniques become more and more sophisticated and able to produce various forms of knowledge, the number of possible applications grows. ML methods tend then to be more precisely specified in terms of the domain knowledge initially required, the control knowledge to be set and the nature of the system output (MLT Consortium, 1993; Kodratoff et al., 1994).

The First International Conference on Information Processing and the Management of Uncertainty (IPMU) was held in 1986 at a time of great debate about the necessity of modelling uncertainty in intelligent systems (which at that time largely meant rule-based expert systems) and the best way of doing so. Whereas the founders of the Conference on Uncertainty in Artificial Intelligence (UAI) in the United States set out with the aim of promoting the use of probability, the organisers of IPMU chose a diametrically opposed course. Though there were a few papers on probability at IPMU '86, the main focus was on alternative methods, primarily those based upon fuzzy sets. Though subsequent conferences have seen greater mix of papers, IPMU remains largely non-probabilistic with the result that the bulk of the participants come from Europe rather than the United States (despite the large amount of work on uncertainty, and especially probability, that is carried out in the US) making IPMU something of a counterpoint to UAT. The difference in participation is exacerbated by the location—whilst the UAI remains in North America, IPMU alternates between Paris and other cities in Europe, including Urbino in 1988 and Palma in 1992.

The representation of parts of legislation in logic, successively implemented in the language of logic programming and managed by Prolog interpreters, has by now existed for more than ten years. The first and most well-known projects were those by the Logic Programming Group of Imperial College of London which, in 1985, formalized the British Nationality Act (Sergot et al., 1986; Sergot, 1990). Other projects followed, for the most part European, including the Italian project, Esplex, developed in Florence (Biagioli et al., 1987), the Dutch project, Prolex, (Walker et al., 1990), the German project born of the collaboration between IBM and the University of Tubingen (Alschwee, Grundrnann, 1986), and the Japanese project, Les-2 (Yoshino, 1986).