A REFERENCE-CONNECTING TECHNIQUE FOR AUTOMATIC INFORMATION CLASSIFICATION AND RETRIEVAL
Abstract
A recent study of command information flow associated with the Dominican Republic coup of April-May 1965 introduced an analytical tool for identifying deficiencies in the flow and use of information which appears to have considerable potential as a general technique for information retrieval. Naval messages are associated with each other through their formal references in a manner analogous to the concept of 'joining' as defined in a recent publication in directed graph theory by Harry, Norman, and Cartwright (Structural Models). 'Reference-connected sets' are then constructed from message traffic dealing with the coup and are found to uniquely identify operational events during the crisis. Such sets (and subsets if further refinement is desired) can be obtained in real-time and a very simple method is demonstrated which automatically classifies messages as they enter the system. This technique, if applied to a library system, avoids both the problem of describing the subject covered in a document and the problem of integrating new subject matter into a predetermined classification code. It is user-oriented system which combines static coupling methods used in the Technical Information Project at MIT and converts them into a synamic process for both classification and retrieval. Thus, it concentrates on the evolution of the subject as represented by referenced documents upon which any 'new' information is based and it has the additional advantage of reflecting the information flow among scientists in any field.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 10, 1967
- Accession Number
- AD0654190
Entities
People
- M. J. Greene
Organizations
- Center for Naval Analyses