The KADS methodology for developing knowledge-based systems attempted to resolve this problem by suggesting that knowledge should be analysed on several different levels simultaneously: the domain level, the inference level and the task level. The development of a set of generic models which serve as a template for the inference level of knowledge has proved extremely useful, with the result that KADS is now the most widely used methodology for KBS development in Europe. CommonKADS, the recent successor to KADS, has extended and refined the recommended representations for each level, so that CommonKADS now provides a comprehensive suite of representations for the analysis of knowledge. In particular, CommonKADS has defined a set of ontological primitives with which domain level knowledge can be analysed.
CommonKADS aims to provide a detailed and re-usable approach to the analysis of acquired knowledge. It is therefore important to understand how the results of various knowledge elicitation techniques should be mapped onto CommonKADS models; however, there is currently little understanding of how the results of structured approaches to knowledge elicitation, such as card sorting or the repertory grid, could be mapped to other formalisms.
Integration between structured knowledge elicitation techniques and the CommonKADS modelling methods is being developed using a hypertext and diagram-based toolkit, known as TOPKAT (The Open Practical Knowledge Acquisition Toolkit). TOPKAT includes support for knowledge elicitation techniques (transcript analysis, laddered grids, card sorting and repertory grids), support for developing the CommonKADS model of expertise, and hyperlinks between different representations. A particularly useful feature is the ability to translate from CommonKADS to the knowledge elicitation tools; this allows one knowledge elicitation technique to generate input for another.
TOPKAT is implemented in AIAI 's hypertext and diagramming tool HARDY and in CLIPS; it runs on machines supporting either X Windows or Microsoft Windows. HARDY provides the ability to create node & link diagrams which can be hyperlinked to other diagrams or to fragments of text; it also provides an interface to CLIPS, allowing CLIPS functions to automate much of the functionality of HARDY.