Inheritance generalized to fuzzy regularity

Inheritance is commonly, exploited by computer programs that deal with hierarchic knowledge and is one instance of the regularity described in this work. Regularity formalizes the observation that hierarchical relationships between concepts reflect relationships between the concepts´ properties, and vice versa. Based on the regularity in a knowledge base, an expansion procedure can compute what the attributes of concepts should be when those attributes haven´t been previously discovered. In tangled hierarchies where a concept has several parents, a true-false model of property values cannot capture the complexity of the relationships among property values. Fuzzy representations of regularity and expansion correct this failure. The fuzzy sets framework takes into account the normative nature of property values. A theorem is presented that relates a parameter of the fuzzy expansion algorithm to the reliability of expansion. A computer program that implements expansion has been applied to several medical knowledge bases. The experimental results show that expansion is reliable