Technical Report Automatic Stratification of Descriptors

First Edition. Variously paginated. 10 7/8 x 8 3/8 inches. Publisher's white printed wrappers with cut out window to view title. Comb bound. Soiling to the wrappers, internally clean. This example with number '00066' stamped on title page. Reproduced typescript printed on recto only. Comb. Moore School Report No: 64-03, prepared for the Office of Naval Research, Information Sciences Branch, under Contract Nonr 551(40). "Library classification systems in general use today today arrange the descriptors or terms, used to catalog the books, into a hierarchy of relations. At the bottom of the hierarchy more terms can be added and new subdivisions created but the structure above the lowest level remains fixed unless the entire tree of relations is to be reconstructed. The hierarchical assignments are made by professional indexers on the basis of current technology and the presently known relationships among the various spheres of human knowledge. A system is here suggested that would make use of a digital computer in order to constantly revise the classification of the descriptors in terms of their actual usage within the document descriptions themselves.A computer algorithm is defind which accomplishes this partitioning and a program was written and run on the IBM 7090." (abstract) "The research that has culminated in the publication of this paper was initiated in response to a specific problem posed by Dr. Noah S. Prywes, Associate Professor of the Moore School of Electrical Engineering. In early 1961 a complete conception of a Multi-List Information Processing System was being formulated (later in the year to be published as a technical report on project Multilist), and a gap existed in the system between the point where a description (being a set of descriptors) was converted to a number to be decoded by a tree to an associative memory address and the point where this tree could accept it as input. The most promising solution suggested was one which assigned temporary code numbers to the descriptors and allowed 'renaming' of these codes in order to maintain a desired consistency between the two above mentioned points of the system. In the ensuing two years a solution along these lines was found and demonstrated on the IBM 7090 computer, but it was soon recognized that this process of renaming, which produced a so-called exclusive partitioning of a descriptor vocabulary, was of itself a potentially useful, aside from its role int he above mentioned Multilist Information Processing System." (preface) The Moore School was home of the famous Moore School Lectures and was a major research computing hub at the this time - a natural partner to work with. Rare, with only a few entries in Worldcat. PROVENANCE: From the library of Artificial Intelligence Pioneer Ray Solomonoff (unmarked).…