| Description |
Transfer of a knowledge-base between systems is necessitated by the great difficulties encountered in acquiring human expert knowledge for computer expert system utilization. Moreover, testing the robustness of the original knowledge-base in a new environment and expanding the domain of the receiving systems are also additional Incentives for performing a knowledge transfer. This project investigated the transfer of the medical knowledge-base embedded in the INTERNIST-1, a medical diagnostic expert system, onto the HELP, a total hospital information system. Analyses of the structures as well as the algorithms used in both systems were used as the basis for finding a suitable transferring scheme. In the empirical transfer of diseases that involve the gastrointestinal tract, the main behavioral features of the INTERNIST-1 as well as the modularity inherent In the HELP system are preserved. An attempt to reduce the information used In making the diagnosis was carried out by assuming that some information is in the noise level and, therefore, expendable. A different, but equivalent, disease scoring algorithm prompted a new look Into the role of the IMPORTANCE concept as it plays in the scoring process of the INTERNIST-1. A more comprehensive index was developed to control the ASKING process. The success of the transfer was measured by the ease of the transfer and by comparing the performance of the hybrid prototype, i.e., the HELP version of the INTERNIST-1, to that of the INTERNIST-1 itself. |
