Extensions to the STAIRS Study—Empirical Evidence for the Hypothesised Ineffectiveness of Boolean Queries in Large Full-Text Databases

  • Authors:
  • Eero Sormunen

  • Affiliations:
  • Department of Information Studies, University of Tampere, Finland. eero.sormunen@uta.fi

  • Venue:
  • Information Retrieval
  • Year:
  • 2001

Quantified Score

Hi-index 0.01

Visualization

Abstract

The STAIRS study conducted by Blair and Maron in the mid-80's is a milestone in the history of IR evaluation. Blair and Maron made strong conclusion about the inadequacy of free-text searching large databases, and their study has been widely referred in the literature to justify the problems of effectiveness in IR systems. However, some critics of the study have plausibly pointed out that the ineffectiveness conclusions were not solidly based on empirical data.This paper introduces a new theoretical and empirical approach to study the problems of high recall searching in large databases and reports the results of a case experiment. The findings verify some of the hypothetical conclusions introduced in the STAIRS study, and expands the picture of falling performance. It is shown that low precision in high recall searching is unavoidable in exact-match Boolean searching since even major concepts are often expressed implicitly in relevant documents. The author suggests that the problem could be reduced in facet-based best-match searching.