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. 2 2 identified in response to the original query, and [OCRerr] and S are the corresponding subsets identified in response to the modified query, [OCRerr] 1 2 [OCRerr] 1 2 then, for the second iteration, the sets [OCRerr] = R U[OCRerr], and .5 = S [OCRerr]S are available for the optimization algorithm. [OCRerr]asically, two alternatives are possible: the optimal vector to differentiate documents in R from documents in S may be used as a perturbation of the user1s original query; or this vector may be used as a perturbation of the query resulting from the first iteration. In practice this could be left for the user to decide, depending on his interpretation of the output from the first iteration. In the general case the expression for the nth modified query in which all modifications are made to the original query can be written: =[OCRerr]1q[OCRerr] + [OCRerr]2 hi $r[OCRerr][OCRerr]> ;-5[OCRerr][OCRerr] 1 1 ([OCRerr].14) where = n iT n n2 = n(s[OCRerr]), and[OCRerr] and[OCRerr] are ½) [OCRerr], S = Q) S , n = n([OCRerr]T), 1 2 i=1 i=1 weighting coefficients. If each modification is made to the preceding resultant query image, the iteration formula becomes: +[OCRerr]2 Fnli I r[OCRerr]i[OCRerr]nl2 I r.[OCRerr]RT s.E 5Th 1 1 ([OCRerr].15)