ISR10 Search Request Formulation chapter Joseph John Rocchio Harvard University Gerard Salton Use, reproduction, or publication, in whole or in part, is permitted for any purpose of the United States Government. case under the index representation assumed) results in a restriction 0£ the range of the angular distance to O[OCRerr]< [OCRerr]j <900,,corr'esponding to a correlation range of 1 > 0. It is now necessary to postulate an explicit objective for any given retrieval operation. And corresponding to any search request rep- resentation,. )q[OCRerr], it is assumed that there exists a non-empty subset DR (DR C D) of the set 0£ reference' document tokens D. This subset, DR, is that'set of document index images.which corresponds to documents in the collection[OCRerr] relevant to the search request. As relevance must be subjectively defined, the specification of the `subset, D11, must be made outside the context of the' retrieval system. It is then assumed that the information needs `of the user can be satisfied by the content of those documents whose index images `are contained in D The case in which D is empty, i.e. when there are nd useful refer'ences in the R source' collection, will be considered separately. The identification o'f the subset D is the goal of[OCRerr] retrieval. R Since `the query-document matching function results in an ordering of `the source collection, an ideal[OCRerr]search r'eq'uest can. be defined as one which induces a ranking on `the elements of D such that all members of the set: D ar R e ranked above (have a higher corre'lation) all other elements of D. [OCRerr]ote that in this definition any degree of relevance or ordering among' the members' of the subset D with re'spect' to their value to the user is R ignored. Since' relevance is a subjective, attribute of a given search request-document collection[OCRerr] pair, determined in theory by t,he individual