Those responsible for storing and indexing information appear best fitted to handle dissemination and retrieval. An earlier paper describing a computer procedure for the first two is here extended to describe a procedure for the latter two. Fully automatic (no editing or dictionary) assignment of key words to documents is described and examples are given. The techniques are statistical but word counting is not used. The techniques are independent of meaning and therefore of language, but inflected languages such as Russian or German may alter results. A new measure of system performance is proposed based on the num ber of transpositions needed to bring the list of output items into agreement with the users opinion. Essential to this concept is the belief that any information system should present an ordered output and adjust able threshold. Relevance is reserved as a mathematical mea sure. An appendix describes some mathematical ideas useful in system development. Keywords selected by the computer from the above abstract: 08 dictionar, 075 editing, 075 languages, 07 indexing, 07 retrieval, 05 language, 03 mathemati.
National Research Council of Canada. Radio and Electrical Engineering Division
Report (National Research Council of Canada. Radio and Electrical Engineering Division : ERB), no. ERB-859.