IRS13 An Analysis of the Documentation Requests chapter E. M. Keen Harvard University Gerard Salton Use, reproduction, or publication, in whole or in part, is permitted for any purpose of the United States Government. x-26 given to this concept c[OCRerr]ipared with stem, producing increases of 1 to 18 1/2, 2 to 11 1/2, 3 to 25 and 4 to 11 1/2. These three reasons for superiority of the thesaurus process are thought to be typical for other requests alsoe There is also a strong corre- lation of reasons a) and b) between the 11A" and VIB?1 requests, since as Figure 15 shows the concepts `1c[OCRerr]puter11 and 11system11 appear a total 16 times in the `1B" requests, and only 6 times in the "A" requests, thus giving the t1Bt1 requests greater opportunity to benefit from the superior handling of these concepts in the thesaurus. This treatment of the different sets of requests only scratches the surface of the problem and points mainly to some of the factors [OCRerr]nown to be involved. D) The Recognition of In[OCRerr]ortant Request Words The presence of quite specific and important single request words and the problem of giving them a weight in proportion to their importance was noted in part 3B. In order to discover whether increases in the weight of such important words does improve retrieval performance, the 35 requests were examined (without knowledge of search results, relevant documents or concept frequency) to see whether such important concepts could easily be {identified. Seventeen requests were found to possess such important concepts, and each of the concepts was tripled in weight. These decisions are recorded in Appendix B. This simulates a quite feasible requestor rule which would as[OCRerr] for any important concepts to be underlined in the request statement, and which the system could recognize and correspondingly increase in weight by some factor. Six requests in addition to the seventeen were also slightly modified, request Al was divided into two as suggested by part 3C; in request