Improving dictionary based data compression by using previous knowledge and interaction

摘要: State of the art lossless data compressors are very efficient. While it is not possible to prove that they always achieve their theoretical limit (i.e. source entropy), effective performances for specific types often close this limit. If we have already compressed a large number messages in past, then can use previous knowledge increase compression current message and design algorithms efficiently compress decompress given knowledge. By doing fundamental coding theorem substitute entropy with conditional new allows better compression. Moreover, if assume possibility interaction between compressor decompressor exploit source. The price might pay low communication errors. In paper study interactive present experimental results on textual data.