FREE TEXT WORD RETRIEVAL AND SCIENTIST INDEXING: PERFORMANCE PROFILES AND COSTS

Abstract

Comparative performance profiles were determined, by two independent scientific information centres, for on‐line retrieval by means of (a) tree text words and (b) subject indexing codes, from a data base of 4,600 descriptions of on‐going research projects in four broad subject areas. Altogether, thirty‐nine questions that users had previously asked of the Science Information Exchange (SIE) of the Smithsonian Institution, and that were in the four subject areas, were employed by SIE staffscientists, and twelve of these were independently used by staff members of the Biological Sciences Communication Project (BSCP) of the George Washington University. Results for the two studies showed average recall values 30–40% higher, and relevance values 15–20% higher, for subject index code use as compared with text word use. Advantages and disadvantages of both approaches are discussed.