THE EXTENSION OF USERS' LITERATURE AWARENESS AS A MEASURE OF RETRIEVAL PERFORMANCE, AND ITS APPLICATION TO MEDLARS
- 1 February 1971
- journal article
- Published by Emerald Publishing in Journal of Documentation
- Vol. 27 (2) , 125-135
- https://doi.org/10.1108/eb026513
Abstract
The performance of a retrieval system with a file of only a few hundred references can be measured by assessing the relevance of each reference to each of a number of queries. A suitable measure of retrieval performance is then the Recall ratio—the fraction of the relevant references that are retrieved by the system. When the file is large this method of measuring performance is not practicable, and Recall cannot be measured although it can be estimated. A number of estimation procedures are examined and found unsatisfactory. Another measure of retrieval performance is the Extension ratio which is approximately the ratio of the quantity of known relevant references before and after putting a query to the retrieval system. The properties of this measure are examined and it is applied to 315 MEDLARS searches.Keywords
This publication has 8 references indexed in Scilit:
- A methodology for test and evaluation of information retrieval systemsPublished by Elsevier ,2002
- A COMPUTER‐AIDED INFORMATION SERVICE FOR NUCLEAR SCIENCE AND TECHNOLOGYJournal of Documentation, 1966
- The evaluation of automatic retrieval procedures— selected test results using the SMART systemAmerican Documentation, 1965
- The Cranfield Hypotheses [with Comment]The Library Quarterly, 1965
- The Evidence Underlying the Cranfield ResultsThe Library Quarterly, 1965
- Testing indexes and index language devices: The ASLIB cranfield projectAmerican Documentation, 1964
- Information Retrieval SystemsScience, 1963
- Evaluation of the performance of an information‐retrieval system by modified mooers planAmerican Documentation, 1963