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. 0 for example, mig'Wb be to regulate the number of non-zero components of the modified query on the basis of the degree of overlap of a component among the relevant documerYs ident[OCRerr]fi[OCRerr] y the user. [OCRerr]here are a number of additional variations to this [OCRerr] relation which might be [OCRerr] [OCRerr]he modification process described above for generating q1 from q0 is amenable to iteration and therefore can be written in the general form: - ii = S ) ([OCRerr].io) i i where is[OCRerr]the ith query of a sequence, and [OCRerr] and S are the relevant and nonrelevant subsets, respectively, identified in response to retrieval with query [OCRerr] It is expected that the rate of convergence of such a sequence to a near optimal query will be rapid enough to make the process economical; however, this is to be investigated experimentally. In any case, the convergence rate can be estimated by the user, since it is reflected in the stability of the retrieved out- put. The user's original query serves to identify a region in the index space which' should contain relevant dQcuments. Since he has no detailed knowledge about the characteristics of the document image's in the store[OCRerr], it is unlikely that the vector image of his query is' optimally located. [OCRerr] identifying relevant' documents in the'region, the user provides the system with sufficient information to attempt to