Potentials and problems of the classification space method

Abstract

A technique for representing human knowledge and judgments developed by Ossorio 1 is described. A matrix of judgments obtained from human experts is factor analyzed, and the resulting factors are used to define an n-dimensional space. Items may be assigned positions in the space by a weighted sum of their component factors. Items close to one another geometrically should be close conceptually. Is a significant amount of information lost in the factor analysis? How stable is the technique? Does the factor analysis extract "conceptually significant" factors? The technique is "humanistic" in approach, and there are ways to determine how well objects are represented in the space. Systems using this method have been promising in information retrieval and medical diagnosis.