A Lifelike Model for Associative Relevance

This paper deals with the general problem of association in artificial intelligence. The suggested approach is based on a multidimensional information space that organizes itself as a large number of elementary associations are supplied. The space contains such entities as words, technical terms, symbols, phrases, and proper names, as may be appropriate for information retrieval, language processing, problem solving, etc. A set-theoretic transformation of the space is applied. The resulting system is capable of responding with whatever entity is most relevant to the entities selected for inquiry. There is no sequential search. Life like characteristics include parallel processing, equipotential memory, tolerance to malfunctions and inexact inputs, random connection and sharing of logical elements, and adaptability. Statistical performance is briefly described, together with the results of small-scale computer simulations. A tentative hard-ware design is outlined, based on MOS techniques currently under development.