An analysis of controlled vocabulary and free text search statements in online searches

Abstract

A series of studies was performed in late 1978 and early 1979 by the ERIC Clearinghouse on Information Resources Special Project staff and consultants to discover how ERIC online searchers were using free and controlled vocabulary in their search statements. Over 40 experienced ERIC online searchers were interviewed and more than half of these volunteered to show us their activity online. Over 650 search traces from 18 different ERIC online searchers using three different retrieval systems were analyzed. Findings show that controlled vocabulary terms are used in 68% of ERIC searches. Controlled vocabulary searching is higher if only terms which resulted in offline printed output are considered. The use of free text was more carefully scrutinized to see if we could discover ways to improve the ERIC database and its online access, since free text searchers have few online search aids besides the basic index with postings information. The vocabulary in 165 free text search statements was analyzed to determine whether ERIC descriptors and identifiers were incapable of representing these concepts. Information for every free text statement was sought on the following questions: (1) Is it a descriptor? (2) Is it cross‐referenced to an accepted term? (3) Is it a variant of a descriptor? (4) Is it expressible using descriptors? (5) Is it expressible using identifiers? (6) Is it only expressible using free text? Five tables summarize the output results of our comparative analysis of free text and controlled vocabulary searching. Six categories of search concepts were discovered for which searchers could only input free text terms or phrases. Online searching tests comparing free and controlled vocabulary were performed using six search topics. The free text search formulations had higher recall and controlled vocabulary formulations had higher precision. Search objective (high recall or high precision), we concluded, should dictate which formulation to use. All our findings helped point to the need for new searching aids for online ERIC searchers. Four displays exhibit what these might be; for example, an on‐line rotated descriptor or identifier display, and some linkage between free text and the Thesaurus. Responsible agents for effecting changes in searching ERIC online are identified and suggestions for improvements are addressed to these three agents.