Indexing and abstracting by association

Abstract

Abstract This article discusses the possibility of exploiting the statistics of word co‐occurrence in text for purposes of document retrieval. Co‐occurrence is defined and related to the mental processes of authors and readers; several means of quantitative measurement of word co‐occurrence are then scrutinized. It is shown that the most strongly co‐occurring word pairs, which are therefore “associated” in a statistical sense, can be represented in the form of an “association map.” The last half of the article presents two modes of use of association maps in literature searching.