Prozessbasiertes Knowledge engineering in medizinischen Problemdomänen

Knowledge elicitation and knowledge modelling have always played an important role within the context of Knowledge Engineering and have as such been named the „bottleneck of system development“. However the phase of elicitation and representation of knowledge is highly responsible for the quality of knowledge-intensive systems being developed. Knowledge-intensive and complex problem domains for which a system needs to be developed are characterized by the mixture of various problem solving methods, a lack of completeness and certainty of validity, and the high numbers of entities. Decision-making processes belong to the knowledge-intensive problem domains. Recent model-based methods of Knowledge Engineering have used templates in knowledge modelling for problem solving methods. Hereby a pre-classification of the problem domain is implicated. Knowledge outside this classification is not recorded in this modelling. The conceptualisation of the domain does not take place independently from the templates. Therefore a reuse of application ontologies is not possible. The new approach introduced within is the process-based Knowledge Engineering. The original decision-making processes of experts are identified and analysed according to the said procedural model. Domain entities are projected on ontological structures regardless of their role in the decision-making process. The representation of the concepts on ontological structures allows a test of validity for concepts and values for problem domains as well as a flexible adaptation to existing information systems. For the presentation of the knowledge there are three levels of representation introduced here: the Konzeptuelle Ebene (conceptual levels of representation), implementierbare Ebene (implementable levels of representation) and Computational Form. The modelling of knowledge takes place on the conceptual level of representation. The entities of a domain and their relations are independently described here. The modelling of knowledge is supported by the said procedural model. The procedural model consists of scenarios employed to analyse separate problem solving processes and displays concepts and knowledge of a domain onto ontological structures and elements of models. In that context ontological structures constitute the link between the entities of the problem domain and the elements of models. The elements of models on the conceptual level of representation have a representation on the implementable level, which is transferred to the Computational Form. To support this method on a technical level the OUZO-tool has been developed. It supports the modelling of knowledge with an editor and directly processes the Computational Form.