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. vil-il b) the size of each query cluster; c) the criterion for an acceptable query cluster. 2) the document collection is partitioned into a set of associated and non-associated documents based on the query clusters in step 1); the fozmtion of subsets of associated documents for each given query cluster can take place in one of three ways: a) the associated subset of documents for the given query cluster is formed by associating documents which correlate highly with a query contained in the given query cluster; b) the associated subset of documents for the given query cluster is formed by associating documents which correlate highly with the centroid vector of the given query cluster c) the associated subset of documents for the given query cluster is formed by associating documents which are judged by a user relevant to his query contained in the given query cluster. The size of the subsets of associated documents depends on what is meant by !Y[OCRerr][OCRerr]gh[OCRerr]y correlated1'; the size can be determined in one of two ways: a) the size depends on the number of queries contained in the given query cluster in the sense that the greater the number of queries contained in a query cluster the greater the number of documents that are associated with a query cluster (this method of determining the size of the associated categories is rationalized by the e[OCRerr]ectation that certain areas of information will more often contain relevant documents for the query, so that these information areas should be larger); b) the size depends on the density of the documents which surround the query in the n-dimensional space; that is, the higher the correlation of the documents with the query the more documents are associated with the query.