ISR11 A Modified Two-Level Search Algorithm Using Request Clustering chapter V. R. Lesser Harvard University Gerard Salton Use, reproduction, or publication, in whole or in part, is permitted for any purpose of the United States Government. vii-i8 triplets are then constructed: ([OCRerr], 0.2,3), ... , ([OCRerr], 0.2, 12), ([OCRerr], 0.3, 3), .., (q, 0.6, 9), (q, 0.6, 12). For each search request which is satisfied, the following data are then obtained: 1) M (q, C, n) equal to the minimum number of documents which are scahned by the given search scheme in order to "satisfy" the request; 2) P (q, c, n) equal to the percentage of documents retrieved by the given search scheme contained in the first n documents retrieved by a full search of the document collection; 5) R ([OCRerr]q, c, n) equal to the number of relevant documents contained in the set of documents retrieved by the given search scheme over the number of relevant documents contained in the first n documents retrieved by a full search of the document collection. Let [OCRerr] = {q1, [OCRerr], ... , [OCRerr]J be defined as the given collection of n test queries, and let [OCRerr]co,no = [OCRerr][OCRerr]l' [OCRerr] ` [OCRerr] no[OCRerr] be defined as the set of all queries such that [OCRerr]q. £ C [OCRerr] 3 co,no implies that the request (q.,co,no) is "satisfied"; the data 3 produced from all request[OCRerr] which were satisfied is then condensed in the following manner: let Kco,no M (co, no) = i=l M(qi. Co,No) 3, Kco,no Kco,no P (co, no) = j=l , Co,No) Kco,no Kco,no R (co, no) = j=l R(~i~~co~No) Kco,no