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. ~-2O In a theoretical framework, the request optimization process focuses on the power of the in[OCRerr]ex transformation to distingnish sets of associated documents within the store by eliminating variances due to particular query formulation. In an operating context, relevance feedback[OCRerr]provides a technique whereby the system user can extract the full power of the index transformation to his retrieval problem, at the cost of iteration (possibly on a sample collection from a large document store..) 5. The Case of [OCRerr]o Relevant Documents The definition of an optimal search request assumed the exist- ence of[OCRerr]a nonempty set of documents relevant to each user?s search request. The relevance feedback query optimization algorithm developed from the definition[OCRerr]as&umes that in response to the retrieval output generated by an initial query, feedback is received identifying bot.[OCRerr]h relevant and nonrelevant documents. Consider now the case in which either there are no relevant documents in the collection or none are identified by the user[OCRerr]response to [OCRerr]the initial retrieval operation. In this case the user is faced with a certain degree of uncertainty. If he is interested in ascertaining that there are in fact no useful documents in the collection, one possibility open to him is as follows: he[OCRerr]may rephrase his search request and resubmit it. The relevance feed- back query modification algorithm, when implemented[OCRerr]with no relevant documents identified, will provide[OCRerr]just the kind of adjustment to the original query which is[OCRerr]use'ful[OCRerr]in such a case. The modified query