Compressing Inverted Files

摘要: Research into inverted file compression has focused on ratio—how small the indexes can be. Compression ratio is important for fast interactive searching. It taken as read, smaller index, faster search. The premise “smaller better” may not be true. To truly build it often necessary to forfeit compression. For lists consisting of only 128 occurrences add overhead. Perhaps list could stored in bytes place words, but must still disk. If minimum disk sector read size 512 and word 4 bytes, then both compressed raw postings would require one seek read. A less efficient technique increase size, decrease load/decompress time, thereby increasing throughput. Examined here are five techniques, Golomb, Elias gamma, delta, Variable Byte Encoding Binary Interpolative Coding. The effect time all measured decompression time. quantitative measure throughput developed performance each method determined.