IRE
Information Retrieval Experiment
Laboratory tests of manual systems
chapter
E. Michael Keen
Butterworth & Company
Karen Sparck Jones
All rights reserved. No part of this publication may be reproduced
or transmitted in any form or by any means, including photocopying
and recording, without the written permission of the copyright holder,
application for which should be addressed to the Publishers. Such
written permission must also be obtained before any part of this
publication is stored in a retrieval system of any nature.
Test types 141
If these results suggest that manual systems have lower recall capability it
should be remembered that the machine searches produce high recall at such
low precision levels that in practice searchers might not tolerate it. If the
bc
8C
0
a
L
a
U
x
6C x
x
4C
2C
20 40 60 80 ½oo
Precision ratio (0I[OCRerr])
Figure 8.2. Retrieval comparison from SMART project, taken
from Keen18. x x manual, KWIC index to abstracts;
automated, abstracts with thesaurus
bc
8C
6C
4c
2C
0
a
L
0
U
Cr
x
x
x
x
x
20 40 60 80 100
Precision ratio (o/[OCRerr])
Figure 8.3. Retrieval comparison of ISILT and Sparck Jones19.
x x manual post-coordinate, uncontrolled language, ISILT;
0 0 automated keyword, run Keen 8001g Ti, Sparck Jones.