Using an intermediate knowledge representation for prototyping

Summary form only given. Knowledge acquisition has been considered a bottleneck in the process of developing expert systems (ESs). Intermediate knowledge representations (IKRs) emerge as a bridge between the expert's knowledge and the compiled knowledge that is to be actually implemented in the knowledge base. The authors present a model for IKRs for diagnosis problems that has been developed as an abstraction effort from practical work carried out in the laboratory. The IKR uses causal networks as its base to capture the inferential knowledge in diagnosis tasks. Some tools have been implemented for translating the IKR to executable forms, allowing testing and validation of the acquired knowledge-that is, building prototypes before implementing the actual ES. This IKR has been successfully used in more than ten ES development projects.<
>