SCM: Structural Contexts Model for Improving Compression in Semistructured Text Databases

Abstract

We describe a compression model for semistructured documents, called Structural Contexts Model, which takes advantage of the context information usually implicit in the structure of the text. The idea is to use a separate semiadaptive model to compress the text that lies inside each different structure type (e.g., different XML tag). The intuition behind the idea is that the distribution of all the texts that belong to a given structure type should be similar, and different from that of other structure types. We test our idea using a word-based Huffman coding, which is the standard for compressing large natural language textual databases, and show that our compression method obtains significant improvements in compression ratios. We also analyze the possibility that storing separate models may not pay off if the distribution of different structure types is not different enough, and present a heuristic to merge models with the aim of minimizing the total size of the compressed database. This technique gives an additional improvement over the plain technique. The comparison against existing prototypes shows that our method is a competitive choice for compressed text databases. Finally, we show how to apply SCM over text chunks, which allows one to adjust the different word frequencies as they change across the text collection.

This work was partially supported by CYTED VII.19 RIBIDI project (all authors) and Fondecyt Project 1-020831 (second author).