A SYSTEM FOR AUTOMATIC CLASSIFICATION OF SCIENTIFIC LITERATURE

Abstract

computer-based system for automatically classifying scientific articles is described. The unique feature of this system is that its structure is completely determined by citation patterns in the Science Citation Index. These citation patterns give rise to clusters or clumps of cited papers which correspond, in turn, to clusters of citing papers. Classificatioll headings for each cluster  are determined by examination of high frequency word pairs drawn from the titles of the citing papers. Classification of a New article is performed automatically by determining what clusters it cites and assigning appropriate (weighted) subject headings to it. The system will permit updating of the classification scheme Oil wilh annual basis, and the incorporation of new headings and deletion of old ones.