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-6 SMART tests. It is surprising to note that the five requests quoted do perform quite well in the retrieval runs made, and of the total of 13 relevant documents involved only 3 receive consistently poor rank positions (below 15) on all search options. E) Difficult Requests Of considerable interest in the analysis of a system such as SMART is the identification of requests that may be quite reasonable in themselves, but that nevertheless create problems due to some system weakness. Six examples are given. Request [OCRerr] contains the foUowing negative statement: "...As opposed to references or entire articles themselves ... ". SMART cannot recognize the significance of the negation, and a search will be made for the ideas as stated. U[OCRerr][OCRerr]ess rules to recognize negative statements can be added to the system, users or request preparers must be advised to avoid negatives. In request A8, 11other languagestt is a very important part of the request, but the idea of "languages other than Xnglish" is another negative statement which cannot be handled. Even if `1other" were replaced by "foreign", correct matches with relevant documents would be difficult to achieve since a thesaurus concept that linka "foreign" with all possible named languages or countries might work well for this particular request, but would at the same time provide an unhelpful grouping for other requests asking for one language in particular. Request AlO contains the homonym `abstract", here used in the sense of "abstract mathematics" rather than the frequent coliections use in the sense of a sumary of a document. The use of phase recognition would cope with this problem, except that the phrase list in use does not contain the required phrase. A synonym problem also exists, [OCRerr]ecause none of the relevant