Edge-Guided Natural Language Text Compression

摘要

We describe a novel compression technique for natural language text collections which takes advantage of the information provided by edges when a graph is used to model the text. This technique is called edge-guided compression. We propose an algorithm that allows the text to be transformed in agreement with the edge-guided technique in conjunction with the spaceless words transformation. The result of these transformations is a PPM-friendly byte-stream that has to be codified with a PPM family encoder. The comparison with state-of-art compressors shows that our proposal is a competitive choice for medium and large natural language text collections.