ISR10 The Query-Document Matching Function 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. 4-7 Comparison Operation Definition E[OCRerr]uality [OCRerr] a = b a. = b. for i = 1,N 1 1 Vector Difference d = a - b d. = a. - b. for i = [OCRerr] 1 1 1 Ma[OCRerr]itude of the [OCRerr]= dl = FZ(ai - b[OCRerr])2[OCRerr]21 Vector Difference J -1 a.[OCRerr]b [OCRerr] An[OCRerr][OCRerr]lar Distance = cos [OCRerr]ai lb[OCRerr][OCRerr] Comparison Operations on Vector Represented Operands Table 4.2 carried by these more complex operand structures is the increased cost in the re[OCRerr]uired comparison operations necessary to specify a retrieved subset or to assi[OCRerr]n a value indicative of document relevance. The discussion here will be primarily concerned with vector operands; however, certain of the results derived will be a function not of the operand structures but of the matching function itself, and will, therefore, be applicable to matching functions of the type considered re[OCRerr]ardless of the operands to which they are applied. 2. Stora[OCRerr] Or[OCRerr]anization In principle, an automatic document retrieval system can be characterized independently of any parameters of storage organization. Given a description of the document and query representations and of