Compression of clustered inverted files

Abstract

One way to save memory space in inverted file organizations is to map each address list to a bit-vector and compress it by a suitable compression technique. Eight such techniques are discussed for nonuniformly distributed bit-vectors in this study. Occurrences of clusters with high 1-bit densities are simulated using a n-state bit-vector generating process. Experiments with a real life file are also reported.