ISR10 The Indexing 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. 2-16 cate[OCRerr]ories associated with each 0£ these concepts. 1£ suchan evaluation indicates that one 0£ these alternatives is [OCRerr]ch more likely than the other, the individual term mappings can be rewei[OCRerr]hted conditioned on this finding. Thus the wei[OCRerr]ht 0£ cate[OCRerr]ory 30 due to occurrences 0£ "channel11 might[OCRerr]be increased from k[OCRerr]n[OCRerr]o( to k[OCRerr]n if the total do'cument[OCRerr]encoding indicated that "m[OCRerr]etic storage" was- more probable than 11in£ormation transmission.tl This second encoding is * * context dependent in th&sense that a [OCRerr]lobal interpretation 0£ the overall original index ima[OCRerr]e has been used to modify the term * - transformation rules. The second method 0£ possible optimization which avoids the necessity 0£ adding context dependent complexity to the index. transforma[OCRerr]tion, focuses on the index images 0£ individual documents. Consider a document retrieval system in which it is possible to record e££ectively the re a ct. ions 0£ the users to the responses 0£ the systeme In the course'o£ a retrieval operation the indexing rules 0£ the system are evaluated by users in the sense that some retrieved - documents[OCRerr]are accepted and others are rejected.[OCRerr] To the extent that these user value judgments re£lect the[OCRerr]indexing accuracy of the system, the information they represent can be used to modify the index images Qf the documents in question. One possible[OCRerr] means of implementing such a scheme is by[OCRerr]an index term weight adjustment algorithm. Consider a document d. retrieved in response, to a search request with index image q.. Define the set of terms common t6 dL and q. to be T . If the,user judges d. 1 J c 1