Knowledge acquisition first, modelling later

Current approaches to knowledge acquisition are based on the idea of modelling with the major effort being put into the initial development, hopefully resulting in models that facilitate reuse etc. There are problems with this in that the situated nature of knowledge leads to the domain model being a partial view resulting in maintenance problems. In contrast, the Ripple Down Rules (RDR) approach emphasizes incremental refinement whereby a knowledge base is built up over time by correcting errors as they occur. Such an approach reduces maintenance problems, but does not provide a model of the domain terms the expert uses, their relationships and various abstraction hierarchies which will facilitate reuse. The paper here describes an approach to discovering a conceptual structure in the domain using Formal Concept Analysis after or during incremental development of the knowledge base. We believe that not only does this assist maintenance and facilitate reuse of the knowledge for such purposes as critiquing and explanation, but it may be a more useful way of helping the experts discover and express significant concepts.