On two inference methods for database systems

This paper proposes two inference methods for database systems. One is called the string inference by which the similarity measure between two strings can be derived. The similarity measure is examined on condition that the string can be divided into significant substrings. The other is called the common item inference which finds some kinds of valid attribute values in the noniput attributes, and using these values the other candidate documents are retrieved. A system construction including those two inference methods is also described.