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.