From Real World Problems to Domain Characterisations

Several problems associated with current approaches to knowledge acquisition have been identified by various workers. These include the lack of an explicit theoretical basis to many existing methodologies and the requirement to be able to integrate the analysis, elicitation and representation of knowledge into a coherent domain model. The purpose of this paper is to address some of these issues. Taking knowledge acquisition itself to be primarily a modelling concern, we examine three general kinds of model of this process. Following this we describe an explicit theoretical framework for one of these model types, namely, the characterisation of problem solving domains. The practical application of these ideas are found within an approach called SAAGS which seeks to provide a characterisation of a domain. This approach has and is being applied in a number of applications, and we provide some details of how it works using selected results from an investigation of the industrial application of neural networks. The final section of the paper looks at the scope of the SAAGS approach and suggests how it could complement other knowledge acquisition techniques.