IRE An experiment: search strategy variations in SDI profiles chapter Lynn Evans 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. Experiment 291 Standard tasks To measure the intellectual effort involved in compiling profiles incorporating the different search strategies, the profile compilation procedure was divided into a number of standard tasks and times recorded for completion of each task. It was considered important that, for meaningful comparison of the search strategies, all the profile versions compiled from a particular user statement should use the same basic set of search terms. The first standard task was therefore to produce a list of search terms. The complete list of seven standard tasks, Ti-T7 (with corresponding task completion times t1-t7) was established as follows Ti-from user statement prepare list of search terms using various aids such as INSPEC thesaurus, dictionaries, known relevant documents, etc. T2-arrange the terms of Ti into groups representing the concepts in the original query. T3-assign boolean equations to govern the groups of T2. T4-assign weights subjectively to the individual terms of Ti. T5-arrange groups of T2 in order of their importance to the original query. T[OCRerr]assign weights subjectively to the groups of T2. T7-arrange the terms in the groups of T2 in order of their importance within the group. Figure 14.1 shows the relationship between the various standard tasks and how they lead to the profiles incorporating the different search strategies. The compilation times are then calculated by adding the appropriate task completion times, e.g. t1 for strategy CT, t1 + t4 for TWC, and so on. Strictly there is no compilation time for strategy CRTW since the profile terms were obtained in rather an indirect way. However it seems reasonable to assume that, starting from a user statement of interests, the time taken to produce a list of the 20 most important search terms (effectively strategy CRTW) would not differ very much from that taken to prepare a complete list of all the search terms likely to be useful (strategy CT). On balance the former task could well involve less time. As an approximation we can take both compilation times as being equal to t1. The compilation of profiles incorporating strategy CLB was a completely separate and self-contained exercise. Clearly, with the standard tasks, there is a question as to how interdependent they might be, e.g. in doing task Ti (preparing list of search terms) does one immediately in one's mind start grouping them into concepts, i.e. task T2, and even consider possible boolean equations, i.e. task T3. Also it might reasonably be argued that the `natural' thing to do starting with the user statement is to isolate the concepts first and then expand them to produce the search terms. With these considerations in mind the user statements were first divided at random into two groups A and B. In group A all the standard tasks for a particular user statement were completed consecutively `at one sitting' whereas for group B each task was completed in isolation, i.e. separated in time from the other tasks, so that the memory of doing one task had largely disappeared before the next one was started.