Extracting knowledge from case databases

The FGP machine is a software architecture that uses similarity-based reminding to make the domain knowledge contained in the data explicit, and then brings that knowledge to bear on information retrieval and machine learning tasks. The FGP machine's goal is to use the cases themselves to drive the system. The system should reason on the basis of specific cases and groups of cases, and should therefore be able to cite specific precedents (including precedents that are themselves incompletely understood), to modify its behavior on the basis of every new information-providing transaction, and to subsume the functions of a conventional information-retrieval system. The author explains the model, and then presents test results for a prototype implementation on a diagnosis task.< >