エキスパート・システムによる出版地の鑑定

抄録

An expert system is one of the major applications of artificial intelligence. Several expert systems have been put into practical use, especially in medicine. The author built a system for the identification of the places of publication of European printed books before 1800, using MECS-AI (Medical Education and Consultation System by means of Artificial lntelligence), a general purpose tool for the medical expert system developed by the University of Tokyo Hospital.
ln this system, the knowledge about the localization consists of several compositorial features (signatures, catchwords, etc.), and is represented as production rules (IF ... THEN …) divided into knowledge source subgroups which are provoked only when they become necessary. The uncertainty associated with a rule is given by a certainty factor (CF: numric judgement of plausibility), and the CFs of the rules are combined, in an inference process, to support or to reject the conclusion parts of the rules.
Although it is difficult to judge the utility and efficiency of the system over a short term, the system seems practical and promising. lt is believed that expert systems for other bibliographical studies can be useful tools for research and education not only in bibliography but in related fields.