Hedge trimming and the resurrection of the controlled vocabulary in online searching

Abstract

The traditional role of the controlled vocabulary has been to bring together items scattered by a multiplicity of natural language terms. Therefore, it was considered a recall device. The file structure of current online structure systems now leads to the use of such vocabularies as a precision device. Because of the levels of specificity in the controlled vocabulary, the search for broader concepts means several terms may be ncessary. In this case, the file structure of the retrieval systems requires that for efficient searching the entry points be analyzed and minimized. Complete, rather than partial, entry of controlled terms appears to be inefficient.

Tidsskriftet bør stille strengere krav til litteratursøk

“…Fifty-six involved research, and these are given the most emphasis here. 7,8,[10][11][12][14][15][16]20,23,24,27,30,34,35,[37][38][39]46,48,59,62,66,67, Thirty-one included theoretical discussion, 7,9,12,16,17,[19][20][21][22][24][25][26][27][28][33][34][35][52][53][54][55][56]62,63,68,72,84,88,…”

“…Twelve dealt with recall only, 10,11,62,70,72,76,82,83,90,95,98,100 four considered precision only, 75,78,79,87 and six dealt with time or cost. 20,35,38,84,91,94 All but one of these 20 also considered precision. Other outcomes considered were searches resulting in no hits, 15 novelty, 67 consensus regarding term selection, 30 specificity, 39,88,89 accuracy, 48 relevance, 14 and diagnostic odds ratio.…”

“…They describe this reduction of controlled vocabulary as "trimming the hedge." 84 In another study that explored query expansion from an initial text search based on users' requests, the addition of thesaurus terms increased recall. 27 A subsequent expansion to related terms further increased recall but with a reduction in precision.…”

Heuristics for online information retrieval: a typology and preliminary listing

“…A hedge is a collection of ored terms that cover a topic for which no single term in the controlled vocabulary has sufficient extension [30]. If a concept is searched frequently, i.e., the concept of 'research' [31], its hedge can be stored, recalled, and executed as often as necessary.…”

“…Searching for 'research' in the descriptor field only will retrieve all records assigned any descriptor that contains the word 'research', such as 'operations research', 'media research', 'medical research', 'classroom research', and 'writing research'. Trimming hedges can serve as an extremely efficient method of searching on a set of related descriptors [30].…”