The [public-domain] mg system uses a local Bernoulli … [in the index]. This provides fast decoding with quite acceptable compression … The within- document frequencies … are stored using the .gamma. code … For the small files [ca. 4-15Mbyte], the index consumed … [and the entire retrieval system was nearly 50% the size of the original]; for the two larger files [ca. 120-2000Mbyte TREC], the augmented inverted file … [and the entire retrieval system was less than 40% the size of the original]. [mg uses huffword for text compression.]
Google-1
Google-2
Copyright clearance needed for quotation.